September 2027

Sarah and Marcus met for coffee in San Francisco on a Tuesday morning. Neither had seen the other since the hearings. Marcus had suggested it-a quick catch-up, see how you’re doing, nothing formal.

They met at a café near the Embarcadero. The Bay stretched out beyond the windows, container ships moving slowly toward Oakland. It was the same view Marcus used to have from his office, forty-two floors up. Now he was at street level.

“You still at CalPERS?” Marcus asked.

“Still there. Chief Risk Officer.”

“How’s that going?”

“I spend most of my time explaining to the board why we can’t make promises we made thirty years ago. You?”

“I’m consulting for the FDIC. Helping liquidate funds. It’s steady work.”

A waitress brought coffee. Sarah stirred in cream, watching it swirl into patterns that dissolved as quickly as they formed.

“I got a letter last week,” Sarah said. “From a retired teacher in Fresno. She worked thirty-two years. She was supposed to get four thousand two hundred a month. Now she gets three thousand five hundred seventy.”

“Because of the benefit cuts.”

“Because CalPERS lost six billion dollars investing in companies that couldn’t make money. Her letter was very polite. She said she understood that ‘market conditions’ caused the losses. She wanted to know if her full benefits would be restored someday.”

“What did you tell her?”

“I told her the truth. No. Not in her lifetime.”

They sat in silence for a moment. Outside, a young couple walked past, laughing about something on a phone. They looked carefree. Like people who hadn’t spent eighteen months watching a financial system slowly collapse.

“I’ve been doing the math,” Marcus said. “Sequoia Harbor’s LPs lost sixty-eight cents on every dollar. We managed two-point-two billion. They got back about seven hundred million. That’s a billion and a half in losses.”

“That’s a lot of teachers’ pensions.”

“Yeah.” Marcus looked out at the Bay. “You ever think about that? About the actual people whose money we lost?”

“Every day. I read their emails. I respond to their questions. I explain to sixty-five-year-olds why they need to work three more years because their pension isn’t fully funded.”

“Does it help? Talking to them?”

“No. But someone should. Someone should look them in the eye and admit we failed them.”

Marcus took a sip of coffee. It had gone lukewarm. “We told them. You told your board. I told my partners. Rebecca told her executives.”

“And nobody listened.”

“Or they listened and decided to wait. Wait for someone else to mark down first. Wait to see if the problem fixed itself. Wait until it was too late to do anything except admit the losses.”

“At least we were right.”

Sarah almost laughed. “Were we? I was right about the losses. I was right about the funded ratio dropping. But being right didn’t help the teacher in Fresno. Being right just means I get to explain to retirees why they’re getting less money for the rest of their lives.”

A container ship moved across the water, slow and massive. It would dock in Oakland, unload its cargo, and head back across the Pacific. Simple. Mechanical. Predictable.

Not like financial markets. Not like human decision-making.

“You know what the crazy thing is?” Marcus said. “The fund managers who lost money-half of them are at other firms now. Six-figure consulting gigs. Advisory roles. Some are starting new funds.”

“Because they know how to raise capital. Because they lost money the same way everyone else did.”

“Whereas you and me-”

“-we saw it coming. We warned people. And we get to spend the rest of our careers cleaning up the mess.”

The waitress refilled their coffee. Marcus ordered a croissant. When it arrived, he broke it apart-flaky layers of pastry, butter and air folded over and over until something simple became complicated.

“I’ve been thinking about something,” Marcus said. “About whether any of this matters.”

“The hearings?”

“The hearings. The testimony. The regulations. The disclosure requirements. Does any of it actually change anything?”

Sarah thought about the new rules. About enhanced reporting requirements and concentration limits and standardized valuation methodologies. About all the paperwork that would be filed and reviewed and forgotten.

“It changes the paperwork. Now funds have to disclose cross-collateral arrangements. Pension funds have to use standardized methods. Insurance companies have to report sector exposure.”

“But does it change the incentives?”

That was the question. The only question that mattered.

“No,” Sarah said. “It doesn’t change the incentives.”

Marcus nodded slowly. Like he’d known what she would say but had needed to hear her say it anyway.

They finished their coffee. Outside, the city continued. People walking to meetings. Closing deals. Making investments. The same things that happened before the crisis. The same things that would happen after.

“I should get back,” Marcus said. “Conference call with the FDIC at two.”

“More liquidations?”

“Always more liquidations.”

They stood to leave. As he picked up his bag, Marcus said: “Sarah, can I tell you something? I spent twenty-seven years in finance. Made it to partner at a major fund. I thought I understood how it worked. Then I watched it all fall apart because everyone made the same mistake at the same time. And the thing that bothers me most isn’t that we failed. It’s that we failed exactly the way I knew we would, and I couldn’t stop it.”

“I know.”

“Do you? Because I keep thinking about that teacher in Fresno. About the six million policyholders. About all the people who trusted the system. And we knew the system was broken. And we couldn’t fix it.”

“We tried.”

“We tried. That’s not the same thing.”

He walked toward the door. Sarah stayed at the table for a moment, looking at the remains of the croissant on his plate. Butter and flour and air, folded into complexity.

She thought about AI companies. About technology that worked but business models that didn’t. About everyone knowing the math was wrong and investing anyway. About simple things made complicated until the truth got lost in the layers.

Her phone buzzed. Her assistant: “Board meeting moved to three. They want Q3 projections.”

Sarah paid for the coffee and walked out into the September sunshine. The Bay Bridge glittered in the distance. Container ships moved toward port. The city looked exactly as it had eighteen months ago.

Everything looked the same.

And in a way, everything is the same.

—–

THE END

“Author’s” Note:

This novel is fiction, but the financial mechanisms described are real. Also, every word of this was written by an AI chatbot – (Claude from Anthropic) based on chats and scenarios explored during research on a possible AI bubble in late October 2025.

Even thought the characters are invented, the situation they face is extrapolated from observable trends in AI investment, private credit markets, pension fund management, and insurance company risk exposure.

The pattern is also real: warnings ignored, problems deferred, losses denied until they can’t be denied anymore.

If this story feels plausible, that should concern you.

March 15, 2027

Janet Rodriguez sat in her FDIC office, reviewing the final accounting. Six months of crisis response. Three insurance companies taken over. Eighteen pension funds stabilized. Forty-three private credit funds liquidated.

She’d assembled the numbers into a briefing for Treasury. Thomas Chen was coming in at two to review it.

POLICYHOLDER PROTECTION FUND – FINAL ACCOUNTING

Insurance Companies:
– Companies taken over: 3 (Pacific Life, MetLife subsidiaries, Prudential subsidiaries)
– Policyholders affected: 6.2 million
– Total insured losses: $38 billion
– Recovery from asset liquidation: $22 billion
– Net federal cost: $16 billion
– Average benefit recovery: 88 cents on dollar

Pension Funds:
– Participating pension systems: 18
– Federal matching funds committed: $54 billion (over 5 years)
– Required state contributions: $54 billion (over 5 years)
– Total public cost: $108 billion
– Average funded ratio improvement: 63% to 68%

Private Credit Markets:
– Funds liquidated: 43
– Total LP capital: $143 billion
– Liquidation recovery: $56 billion
– Total LP losses: $87 billion
– Federal support provided: $0

Public Markets:
– S&P 500 peak-to-trough decline: 23%
– Current level: 8% below pre-crisis peak
– Household wealth destroyed: $4.2 trillion

Thomas Chen arrived at exactly 2 PM. He read through the briefing in silence.

“Sixteen billion to protect insurance policyholders. Fifty-four billion for pensions. That’s seventy billion in federal money.”

“To clean up a hundred billion in private losses,” Janet added.

“Could have been worse.”

“Could have been prevented.”

Thomas looked at her. “You really think so?”

Janet pulled up another document. “I’ve been reviewing the timeline. Sarah Chen identified the risk in December 2025. Rebecca Torres ran stress tests in January 2026. Marcus Webb modelled contagion in February. We had nine months of warning.”

“What would you have done differently?”

“Required immediate mark-downs. Mandatory stress testing. Enhanced supervision of any institution with more than fifteen percent exposure to a single sector. Not let everyone wait to see if someone else would blink first.”

“That would have triggered the crisis earlier.”

“It would have triggered honest accounting earlier. The crisis was always going to happen. The question was whether we’d have it when the losses were four billion or when they were a hundred billion.”

Thomas closed his laptop. “You’re probably right. But you’re also describing a politically impossible intervention. We can’t force pension funds to mark down positions that fund managers say are fine. We can’t take over insurance companies that meet capital requirements on paper. We can’t act on predictions, only on facts.”

“The predictions were facts. They had data. Models. Analysis.”

“They had analysis that their bosses ignored. That’s different.”

Janet knew he was right. That’s how it always worked. The people with the data weren’t the people with the power. And the people with the power didn’t want to hear bad news.

After Thomas left, Janet pulled up one more document. The human cost.

INDIVIDUALS AFFECTED:

Insurance policyholders: 6.2 million receiving 88% of promised benefits
Pension beneficiaries: 8.4 million receiving 85% of promised benefits
AI sector employees: 180,000 jobs lost
Retirement age impact: Average delay of 2.3 years
Household wealth impact: $4.2 trillion destroyed

She thought about the numbers. About what they meant. A seventy-two-year-old policyholder getting $2,800 per month instead of $3,200. A teacher working two extra years because their pension wasn’t enough. A fifty-year-old engineer laid off from an AI startup, unemployable because their skills were in a sector that had imploded.

Numbers on a spreadsheet. Lives in the real world.

Her phone rang. Her boss: “Janet, good work on the Pacific Life wind-down. We’re getting ready to announce you’re heading up the new Insurance Supervision division. Bigger team, larger mandate, better office.”

“What’s the mandate?”

“Making sure this doesn’t happen again.”

After she hung up, Janet looked at her briefing. At the seventy billion in federal money. At the hundred billion in losses. At the millions of people affected.

Making sure this doesn’t happen again.

That was the promise they always made. After every crisis. After savings and loans in the eighties. After dot-com in 2000. After housing in 2008.

And it always happened again. Different sector. Different instruments. Same pattern.

Because the system was designed to create bubbles. And regulators could only respond after the bubble popped.

She started drafting the new supervision framework anyway.

Because that’s what you did.

—–

David and Jamie sat in a cafe in Mountain View, having coffee on a Thursday afternoon. They’d been meeting every few weeks since David started at Meta. Not for any particular reason. Just two people who’d been through something and wanted to talk to someone who understood.

“You see the final accounting?” Jamie asked. “Hundred billion in losses.”

“I saw. Artemis was what, a hundred twenty million of that?”

“A rounding error. Still doesn’t make me feel better.”

“Why would it?”

“Because if everyone lost money the same way, maybe it wasn’t my fault specifically.”

David stirred his coffee. “You think it was your fault?”

“I built a company with unit economics that didn’t work. I burned a hundred twenty million dollars. I laid off forty-seven people. Of course it was my fault.”

“But you weren’t wrong about the technology. The medical imaging AI actually worked.”

“The technology worked. The business model didn’t. That’s the part I should have figured out before raising all that capital.”

They sat in silence for a moment. Outside, people walked past. Young employees from the tech companies that dotted the area. Some probably working on AI projects right now. Some probably building things that would fail for the same reasons Artemis failed.

“You ever think about going back?” David asked. “Starting another company?”

“Every day. Then I remember that I have no credibility, no track record of success, and a clear history of burning investor money. Who’s going to fund me?”

“Someone who wants to learn from your mistakes.”

“Nobody wants to learn from mistakes. They want to learn from successes. That’s why all the fund managers who lost money in the AI bubble are starting new funds. They have relationships. They know how to raise capital. It doesn’t matter that they were wrong. It matters that they were wrong the same way everyone else was wrong.”

David thought about that. “Is that why you’re consulting instead of starting over?”

“I’m consulting because I don’t trust my own judgment anymore. I spent four years convinced we’d figured it out. Convinced that we just needed one more funding round, one more customer, one more breakthrough. And I was wrong the entire time. How do I know I won’t make the same mistake again?”

“You could make different mistakes.”

Jamie laughed. “That’s not as reassuring as you think it is.”

They finished their coffee. Talked about other things. Weather, sports, plans for the summer. Normal conversation between two people who’d been through something abnormal.

As they were leaving, Jamie said: “David, can I ask you something? Do you feel guilty?”

“About what?”

“About getting out. About landing at Meta while everyone else from Artemis was scrambling for jobs.”

David had been thinking about this for months. “Yeah. Every day. I see people from the old team on LinkedIn. Some are still unemployed. Some took jobs that are obvious downgrades. And I’m at Meta making three hundred eighty thousand a year.”

“You took a job offer. That’s not a crime.”

“It’s not a crime. But it doesn’t feel fair either.”

“Life’s not fair.”

“I know. Still bothers me.”

They parted ways. Jamie walked to his car, thinking about fairness. About how he’d failed and David had succeeded and neither of them felt good about their outcomes. About how the system rewarded survival more than correctness, luck more than skill, timing more than judgment.

About how he’d built something that could have saved lives but couldn’t make money. And how that might have been the wrong order to do things.

—–

Rebecca Torres sat at her desk, processing the last of the Pacific Life complaints. She’d been doing this for six months. Eight complaints per day, five days per week. She’d handled 1,247 complaints total.

Complaint 1,247 was from an eighty-year-old woman in San Diego whose annuity had been reduced by $430 per month. Rebecca wrote the same response she’d written 1,246 times before:

The FDIC has maximized recovery for all policyholders based on available assets. Your annuity payments will continue at 88% of the originally promised amount…

She hit send.

That was the last one.

Pacific Life was officially wound down. The insurance company that had existed for 153 years was gone. Dissolved. Assets liquidated, policies transferred, complaints processed.

Her phone rang. Robert Chew, her old boss.

“Rebecca, I saw that you finished the wind-down. Congratulations.”

“Thanks.”

“What’s next for you?”

Rebecca had been thinking about this. She had a standing offer from the FDIC to stay on permanently. She had interview requests from insurance companies that wanted someone who’d been through a crisis. She had an offer from USC to teach risk management.

“I think I’m going to teach.”

“Really? You’d leave the FDIC?”

“I’ve spent six months explaining to people why their retirement income is lower than promised. I’d like to spend the next few years teaching future risk managers how to see problems before they become crises.”

“You think they’ll listen?”

“Some will. Maybe one in twenty. That’s better than none.”

After they hung up, Rebecca packed up her desk. She didn’t have much. A few files. A coffee mug. A photo of her family.

On her way out, she stopped by Janet Rodriguez’s office.

“Wanted to say goodbye. Thanks for the opportunity.”

Janet looked up from her computer. “You heading to USC?”

“Starting in September.”

“What are you going to teach them?”

“I’m going to teach them that risk management isn’t just about models and spreadsheets. It’s about knowing when to ignore your boss and tell the truth anyway. Even when nobody wants to hear it. Even when it costs you your job.”

“That’s not in the curriculum.”

“It should be.”

Rebecca left the FDIC building and drove south toward San Diego. She had six months before classes started. Time to think about what she’d learned. Time to figure out how to explain to twenty-five-year-olds that the system was designed to ignore warnings until it was too late.

But she’d try anyway.

Because someone had to.

—–

Sarah sat in her CalPERS office, reviewing the Q1 2027 projections. The numbers were slightly better than Q4. Markets had partially recovered. The funded ratio had ticked up from 63% to 65%.

Still far from the 72% they’d been at before the crisis. Still far from the 100% that full funding required.

Her phone rang. A number she didn’t recognize.

“Ms. Chen? This is Katherine Wells from Harvard Business School. I’m developing a case study on the AI sector collapse. Would you be willing to participate?”

“What would that involve?”

“An interview about how you identified the risk, how you tried to communicate it, and what lessons you think institutions should learn. Your testimony has become quite well-known in risk management circles.”

“Has it changed anything?”

Professor Wells paused. “Changed? I’m not sure. But it’s being taught. Students are studying your analysis. That’s something.”

After she agreed and hung up, Sarah thought about that. Students studying her analysis. Future risk managers learning about cross-collateralization and concentration risk and the importance of marking down positions honestly.

Would it matter? Would they remember these lessons when they were in positions of power? Or would they do what every previous generation had done-make their own mistakes, ignore their own warnings, create their own crises?

She didn’t know.

But at least there would be a record. At least someone would have tried to teach them.

Her assistant knocked on the door. “The board meeting’s starting in five minutes.”

Sarah gathered her materials. Another board meeting. Another set of projections. Another discussion about when benefits could be restored.

She walked down the hallway, past the photos of CalPERS through the decades. Pictures of ribbon cuttings and retirement ceremonies and happy retirees.

They’d have to take down some of those photos. Replace them with new ones. Photos of a pension fund that had learned expensive lessons about concentration risk and private market valuations.

Or maybe they’d just leave them up. Reminders of better times. Of when the promises had been easier to keep.

She walked into the boardroom and took her seat.

Time to explain, once again, why full funding was still twenty-five years away.

February 10, 2027

Sarah sat at her kitchen table, eating breakfast and scrolling through the Wall Street Journal on her phone. The headline made her stop mid-bite:

“Senate Grills Fund Managers Over AI Losses”

She opened the article:

WASHINGTON-The Senate Banking Committee held contentious hearings Tuesday examining how pension funds, insurance companies, and private equity firms lost more than $100 billion in the collapse of the artificial intelligence sector.

Thomas Werner, representing the American Investment Council, testified that “unprecedented market conditions” led to losses that “no reasonable risk model could have predicted.”

Senator Patricia Mills (D-CA) challenged this assertion, citing testimony from February 2026 by CalPERS analyst Sarah Chen, who warned of systemic risk in AI investments nine months before the sector collapsed.

“Ms. Chen predicted losses of $4 to $6 billion for CalPERS alone,” Mills said. “The actual losses were $6.2 billion. This was not unpredictable. It was predicted.”

Werner acknowledged that “some analysts raised concerns,” but maintained that “the speed and severity of deterioration exceeded even pessimistic projections.”

Sarah stopped reading. Put down her phone. Stared at nothing.

Some analysts raised concerns.

That’s what they’d reduced it to. Some analysts. Not: multiple risk managers at multiple institutions saw this coming and documented it. Not: we ignored specific, quantified warnings. Just: some analysts raised concerns.

Her phone buzzed. Rebecca Torres: “You reading this?”

“Yeah.”

“We’re ‘some analysts.'”

“I noticed.”

“Eighteen months of work. Testimony. Models. Memos. And we’re a footnote.”

Sarah didn’t respond. What was there to say?

Later that morning, she got an email from Marcus. The subject line: “Senate Testimony – February 15.”

Sarah,

Got subpoenaed to testify next week. My lawyers say to keep answers short, don’t volunteer information, don’t speculate. Basically: be as unhelpful as possible while staying technically truthful.

Think I’m going to ignore that advice. If they’re going to make us testify, might as well tell the truth.

Marcus

—–

February 15, 2027. Marcus sat at the witness table in the Senate Banking Committee hearing room. He’d spent three days preparing with lawyers. They’d given him a script: acknowledge limited visibility, cite unprecedented conditions, express regret, commit to learning from the experience.

He’d listened politely and decided to ignore all of it.

Senator Mills was questioning him now.

“Mr. Webb, you were a partner at Sequoia Harbor Capital, correct?”

“Yes.”

“When did you first identify problems with cross-collateralization in AI equipment financing?”

“December 2025.”

“And you warned your partners?”

“Yes. January 2026.”

“What did they say?”

Marcus could follow the script here. He could say “my partners and I had different risk assessments.” He could say “we were monitoring the situation closely.” He could use any of the careful phrases his lawyers had prepared.

Instead he said: “They said I was being too conservative. That the sector would stabilize. That marking down our positions would make us look weak compared to other funds.”

Senator Mills leaned forward. “Mr. Webb, did you consider exiting these investments when you identified the risk?”

This was the question his lawyers had spent the most time on. The one where he was supposed to explain about fiduciary duties and illiquid positions and market conditions.

“Senator, by the time I identified the problem, the losses were already baked in. We just hadn’t admitted them yet.”

The room went quiet.

“Could you elaborate on that?”

“The companies were already burning cash faster than they could generate revenue. The equipment was already cross-collateralized. The intercreditor disputes were already inevitable. We couldn’t exit because there were no buyers. Nobody wanted to purchase positions in companies with unit economics that didn’t work. So we held them, hoping they’d improve, knowing they probably wouldn’t. And when they failed, we lost sixty-eight cents on every dollar our LPs had invested.”

Senator Mills made a note. “Mr. Webb, do you believe this crisis was predictable?”

The lawyer on his right was signalling him to be careful. This was where he was supposed to talk about unprecedented conditions.

“Yes. Multiple people predicted it. Sarah Chen at CalPERS. Rebecca Torres at Pacific Life. Analysts at other institutions. We all saw the same pattern: unsustainable unit economics, concentrated risk, cross collateralization creating contagion risk. The crisis was extremely predictable. We just didn’t want to act on the predictions because acting would have required admitting losses earlier.”

“Why didn’t anyone act?”

“Because everyone was waiting for someone else to move first. Nobody wanted to be the pension fund that marked down while everyone else kept marks high. Nobody wanted to be the fund that suspended redemptions while others stayed open. So we all waited. And while we waited, the losses got worse.”

After his testimony, Marcus stood in the hallway. His lawyer approached, looking unhappy.

“That was not what we prepared.”

“I know.”

“You essentially admitted that your firm ignored risk warnings.”

“We did ignore risk warnings.”

“Mr. Webb, I need to advise you that your testimony today may expose you to legal liability-”

“For telling the truth? That’s not liability, that’s honesty. Maybe someone should try it.”

His lawyer walked away. Marcus stood alone in the hallway, waiting for nothing in particular.

His phone buzzed. Sarah Chen: “Watched your testimony. Thank you.”

He typed back: “For what?”

“For not pretending.”

—–

That evening, Sarah sat in her living room, watching news coverage of the hearings. The segment was brief-three minutes between a story about infrastructure spending and a story about the President’s approval rating.

The anchor summarized: “Senate hearings examined the collapse of the AI investment sector, with fund managers attributing losses to unprecedented market conditions. No criminal referrals were made.”

Three minutes. Eighteen months of crisis. One hundred billion in losses. Six million people affected.

Three minutes.

Her phone rang. Martin Zhao from CalPERS.

“Sarah, the board wants to know if you’re planning to testify.”

“The Senate didn’t subpoena me. I testified last year.”

“Good. We’d prefer to avoid additional attention. The PR team thinks your February 2026 testimony is being characterized as ‘controversial’ rather than ‘prescient.'”

“Martin, I predicted the losses within ten percent accuracy.”

“I know. But the narrative isn’t about accuracy. It’s about not rocking the boat.”

After he hung up, Sarah turned off the TV. She didn’t need to watch anymore. She knew how this would end.

A week of hearings. Stern warnings from senators. Maybe some technical regulation about disclosure requirements. Then everyone would go back to work. Fund managers would find new jobs. Pension funds would rebuild their portfolios. Insurance companies would resume writing policies.

And in five years, or ten years, there would be a new sector. New investment theme. New concentrations of risk that everyone would pile into until it collapsed.

Because that’s what happened. Not because people were stupid. But because the incentives never changed.

You got rewarded for making money the same way everyone else made money. You got punished for being cautious when everyone else was aggressive. You got fired for being right early instead of wrong late.

The system was designed to create bubbles. And then act surprised when they popped.

Sarah opened her laptop and searched for news coverage of Marcus’s testimony. One quote had been pulled out for headlines:

“The losses were already baked in. We just hadn’t admitted them yet.”

At least someone had told the truth.

Even if nobody really wanted to hear it.

December 15, 2026

Sarah Chen sat in her new office on the fourth floor of CalPERS headquarters. It was larger than her old office-corner views, better furniture, a title on the door that said “Chief Risk Officer.” She’d been promoted three weeks ago, right after the federal rescue package passed.

It felt less like a promotion and more like being made captain of a sinking ship.

Her phone rang. Martin Zhao, the Chief Investment Officer.

“Sarah, the board wants updated projections for the Q4 review. Can you have something ready by Friday?”

“What kind of projections?”

“Path to full funding. They want to know when we can restore benefits.”

Sarah pulled up her spreadsheet. She’d been running these numbers for weeks. They never got better.

“Martin, at current contribution rates and assuming seven percent returns-which we’re not getting-we’ll reach full funding in twenty-eight years.”

Silence on the line.

“That’s not a number I can give the board.”

“It’s the only number that’s honest.”

“Can we assume higher returns? Maybe private markets recover-”

“Martin, we just lost six billion dollars in private markets. We’re not going back to thirty percent private allocation. We can’t.”

“Then what do I tell them?”

Sarah looked out her window. The parking lot was full of cars belonging to CalPERS employees-people who themselves were in the pension system, people who would retire someday expecting benefits that were no longer fully funded.

“Tell them we’re implementing the best practices I outlined in my testimony. Tell them we’re improving risk management and diversification. Tell them we’re working to restore full funding as quickly as possible. But don’t tell them when. Because we don’t know.”

After he hung up, Sarah opened her email. One hundred forty-seven unread messages. She started working through them.

From: [member email address]
Subject: Benefit Reduction Question

Dear Ms. Chen,

I retired in July after 32 years as a teacher. I was supposed to receive $4,200 per month. My first payment was $3,570. Is this temporary? When will my full benefits be restored?

Thank you,
Margaret Silva
Fresno

Sarah had received sixty-three similar emails in the past month. She’d crafted a response that was honest without being cruel:

Dear Ms. Silva,

Thank you for your email. I understand your concern about the benefit reduction.

Under California Government Code Section 20221, CalPERS is required to implement benefit restrictions when our funded ratio falls below certain thresholds. Due to investment losses in Q2 and Q3 2026, our funded ratio is currently 63%. This requires reducing new retiree benefits to 85% of the full formula.

The reduction will remain in effect until CalPERS returns to full funding. We are working diligently to restore full benefits, but I cannot provide a specific timeline.

I deeply regret the impact this has on you and other members. You fulfilled your service commitment to California. You deserve full benefits. I wish I had better news.

Sincerely,
Sarah Chen
Chief Risk Officer

She sent it and moved to the next email. And the next. And the next.

By evening, she’d responded to thirty-eight members. She had sixty-seven more to go.

Her phone buzzed. A text from Rebecca Torres: “How you holding up?”

Sarah typed back: “Explaining to retirees why they’re getting 85% of promised benefits. You?”

“Explaining to insurance policyholders why their annuities are being reduced. Same conversation, different institution.”

“Want to get dinner?”

“Can’t. Working late. FDIC wants Pacific Life fully wound down by year-end. I’m processing 2,800 policyholder complaints.”

Sarah put her phone down. She looked at the title on her door. Chief Risk Officer. The person responsible for making sure this never happened again.

Except it would happen again. Different sector, different instruments, same pattern. Because the fundamental problem wasn’t risk management. It was incentives. And incentives didn’t change just because you wrote better policies.

She opened her laptop and started drafting a memo to the board. Not about projections or timelines. About something more fundamental.

MEMORANDUM

TO: CalPERS Board of Directors
FROM: Sarah Chen, Chief Risk Officer
DATE: December 15, 2026
RE: Structural Changes to Investment Policy

Summary: The AI sector collapse revealed systemic weaknesses in how CalPERS manages risk. This memo proposes structural changes to prevent similar losses in future market cycles.

She wrote for two hours. When she finished, she had a ten-page document outlining new concentration limits, enhanced due diligence requirements, and independent valuation processes.

She knew the board would approve maybe half of it. Fund managers would lobby against the rest. The changes that survived would be watered down through “practical implementation considerations.”

But she had to try.

Because that’s what you did. You saw the problems. You documented them. You proposed solutions. Even though they probably wouldn’t be enough. Even though the next crisis would find a different weakness to exploit.

You did it anyway.

—–

Marcus Webb sat in his apartment in the Marina, looking at his severance documents. Sequoia Harbor Capital had officially dissolved on December 1st. The liquidation had yielded exactly $704 million for limited partners who’d invested $2.2 billion.

Thirty-two cents on the dollar.

His phone rang. Janet Rodriguez from the FDIC.

“Marcus, I know this is short notice, but we’re looking for someone with your background. Consulting role. Helping us wind down private credit funds that are in resolution.”

“How many funds?”

“Seventeen in our current pipeline. Probably more in Q1.”

“What’s the pay?”

“Two hundred an hour. As much work as you want.”

Marcus did the math. Forty hours a week at two hundred an hour was four hundred thousand a year. Not partner-level money, but more than most people made. And the work was steady.

“What does the job involve?”

“You’d analyze portfolio companies, determine asset values, help us figure out recovery rates for creditors. Basically what you did at Sequoia Harbor, but for us instead of LPs.”

“So I’d be calculating how much money sophisticated investors are going to lose.”

“That’s one way to put it.”

Marcus thought about it. He could say no. He could retire at fifty-three with the money he’d saved over twenty-seven years. Not enough to maintain his old lifestyle, but enough to live comfortably.

Or he could take the job. Use what he’d learned. Help clean up the mess.

“When do you need an answer?”

“Friday.”

“I’ll let you know.”

After she hung up, Marcus walked to his window. The view from his apartment was nothing like the view from his old office. No Bay Bridge, no soaring glass towers. Just street level. Pedestrians and cars and normal life.

He thought about his former partners. Tom Hendricks had landed at another fund. Karen Reyes was general counsel at a tech startup. Most of the team had found jobs within six months.

Everyone except Marcus. Because Marcus was the one who’d warned them. The one who’d built the contagion model. The one who’d testified before the Senate.

You didn’t hire the person who’d been right. You hired the person who’d been wrong the same way everyone else was wrong.

His phone buzzed. Tom Hendricks: “Drink this week? Catch up?”

Marcus stared at the message. They hadn’t spoken since the liquidation. What would they even talk about? The good old days when they’d thought they were smart?

He typed back: “Sure. Thursday?”

Because what else was there to do?

—–

David Huang sat in his Meta office in Menlo Park, reviewing code for a project he didn’t believe in. The team was building AI agents for customer service. The unit economics were bad. Every interaction cost Meta $0.40 in compute. They were charging customers $0.15.

His manager walked by: “How’s it coming?”

“Code works. Economics don’t.”

“That’s a future problem. Right now we need to ship.”

After his manager left, David pulled up LinkedIn. He’d been at Meta for six months. The pay was excellent. The work was fine. But every project felt like Artemis all over again-technically impressive, economically unsustainable.

His phone buzzed. Jamie Morrison, his old CEO.

“You free for lunch?”

They met at a Thai place in Palo Alto. Jamie looked different. Less stressed. He’d lost the hunted expression he’d had in the final months at Artemis.

“You working?” David asked.

“Consulting. I’m advising a hospital system on realistic AI integration. How to use the technology without losing money on every transaction.”

“There’s a market for that?”

“Turns out yeah. After a hundred billion in AI losses, companies want someone who can tell them which use cases actually work economically.”

“What do you tell them?”

“Same thing I should have realized at Artemis. AI is amazing for automation where you’re replacing human labor on repetitive tasks. It’s terrible for anything that requires scaling compute costs faster than revenue.”

David thought about that. “So most of what we built was in the wrong category.”

“Most of what everyone built was in the wrong category. The medical imaging AI actually worked-we just chose a terrible business model. We should have licensed it to existing radiology practices, not tried to sell direct to hospitals.”

“Would that have saved us?”

“No. But it would have meant fewer people losing jobs. Smaller losses. More honest about what we were building.”

After lunch, David walked back to his car. He thought about his Artemis equity. $270,000 on paper, worth exactly zero in reality. Four years of his life. Four years of believing the vision.

His phone showed his Meta RSUs: $627,000 vested over four years. More than he’d ever made at a startup.

But it didn’t feel like winning. It felt like he’d survived and others hadn’t. Like he’d been lucky, not smart.

He drove back to Meta, back to code that worked but economics that didn’t, back to a job that paid well but meant less than he’d hoped.

Survivor’s guilt, they called it.

It felt exactly right.

—–

Rebecca Torres sat at her FDIC desk, processing her two-hundred-sixteenth policyholder complaint of the week.

Complaint ID: PL-2026-8847
Policyholder: Robert Chen, age 72
Issue: Annuity payment reduced from $3,200/month to $2,816/month

Details: I paid into this annuity for thirty years. Pacific Life promised me $3,200 per month for life. Now I’m getting $2,816. Is this legal? Can they do this?

Rebecca typed her response:

Dear Mr. Chen,

When Pacific Life became insolvent, the FDIC took over the company. We are paying claims up to 88 cents on the dollar, which is the maximum recovery based on asset liquidation.

Under state insurance guaranty funds, you are protected up to $250,000 in total benefits. Your annuity falls within this protection.

While I understand this is not the full amount you were promised, the FDIC is working to maximize recovery for all policyholders.

Sincerely,
Rebecca Torres
FDIC Resolution Specialist

She hit send. Moved to the next complaint. And the next.

Her phone rang. It was her old boss from Pacific Life, Robert Chew.

“Rebecca, I wanted to check in. See how you’re doing.”

“I’m doing fine. Busy.”

“I heard you’re handling policyholder complaints.”

“Someone has to.”

“Look, I wanted to say-you were right. About the stress tests. About the risk. I should have listened.”

Rebecca was quiet for a moment. She’d been waiting for this call for eight months. Now that it had come, she didn’t know what to say.

“Thank you for saying that.”

“Does it help? Knowing you were right?”

“Not really. The policyholders are still getting reduced benefits. Pacific Life still failed. Being right just means I get to explain to seventy-two-year-olds why their retirement income is lower than promised.”

“For what it’s worth, the new regulations-the ones that came out of the crisis-they’re based on your stress test methodology. The FDIC is requiring insurance companies to run similar scenarios.”

“Will it prevent the next crisis?”

Robert was quiet. “Probably not. But it might make it less bad.”

After he hung up, Rebecca went back to the complaints. She had eighty-four more to process before she could go home.

This was what vindication looked like. Not a victory parade. Not acknowledgment. Just more work. More explaining. More trying to help people who’d been hurt by a system that nobody had fixed.

—–

Jamie Morrison sat in a Starbucks in Palo Alto, writing his Medium post. He’d been writing it for three weeks. It was up to 4,000 words.

Title: “Why AI Startups Are a Trap: Lessons from Building One That Failed”

He’d published it that morning. By evening, it had 12,000 views. By the next day, 40,000. By the end of the week, 200,000.

People were sharing it. Commenting. Arguing. Some said he was bitter. Some said he was honest. Some said he should have known better.

One comment stood out:

“This is exactly what I’ve been trying to tell my VCs. The unit economics don’t work. They keep saying ‘scale will fix it.’ It won’t. Thank you for writing this.”

Jamie replied:

“Scale doesn’t fix unit economics when your costs grow faster than your revenue. I learned this the hard way. You’re learning it the smart way-by reading instead of experiencing. Good luck.”

His phone rang. A number he didn’t recognize.

“Jamie Morrison? This is Katherine Wells from the Harvard Business School. I’m teaching a case study on AI company failures. Would you be willing to talk to my students about Artemis?”

“What would you want me to say?”

“The truth. What you got right, what you got wrong, and what you’d do differently.”

Jamie thought about it. About the forty-seven employees he’d laid off. About the medical imaging AI that actually worked. About the unit economics that never made sense.

“I can do that.”

“We pay speakers. Not much, but-”

“I don’t want money. I want the students to understand that you can build something technically brilliant and economically doomed. That most people won’t tell them this because admitting it means admitting they failed.”

“That’s exactly what I want them to understand.”

After she hung up, Jamie looked at his Medium post. The view count was still climbing. People were reading it. Learning from his mistakes. Maybe preventing their own.

It didn’t make up for Artemis failing. It didn’t give his employees their equity back. But it was something.

Sometimes being a cautionary tale was the most useful thing you could be.

November 12, 2026

Janet Rodriguez sat in a conference room at Treasury headquarters, listening to officials debate how to stop a financial crisis without calling it a bailout.

“We can’t use TARP-style capital injections,” Thomas Chen from Treasury was saying. “Politically impossible. The last administration spent $700 billion bailing out banks. This administration can’t be seen doing the same thing for insurance companies.”

“So what do we do?” Michael Park from the Fed asked. “Stand by and watch six million policyholders take losses?”

“We find another mechanism. The FDIC has the Deposit Insurance Fund. Can we use something similar for insurance companies?”

Janet spoke up. “There’s already a structure-state guaranty funds. Every state has one. When an insurance company fails, the guaranty fund covers policyholder claims up to certain limits. Usually $250,000 per policy.”

“How much capital do these guaranty funds have?”

“Maybe $10 billion across all states combined.”

“And we need?”

“$40 billion to cover Pacific Life, MetLife, and the three other companies that are approaching insolvency. Plus another $30 billion if the crisis spreads to smaller insurers.”

Thomas whistled. “So the guaranty funds are underfunded by a factor of seven.”

“At least.”

The meeting went on for three hours. Every solution they proposed had fatal flaws:

– Direct capital injections: Politically impossible, looks like bailouts
– Emergency guaranty fund: Would take months to authorize, requires state coordination
– Federal backstop: Might work but sets precedent for future bailouts
– Liquidation: Policyholders take massive losses, triggers more contagion

Finally, someone suggested: “What if we don’t bail out the companies? What if we just guarantee that policyholders don’t take losses?”

Thomas leaned forward. “Explain.”

“The FDIC takes over the failing companies. They liquidate the assets, recover what they can. For any shortfall, Treasury creates a special facility-let’s call it the Policyholder Protection Fund-that covers the gap. It’s not a bailout of the insurance companies. It’s protection for individual policyholders.”

“That’s still $40 billion from Treasury.”

“Yes, but the optics are completely different. We’re not bailing out Wall Street. We’re protecting teachers and retirees and middle-class families who bought insurance in good faith.”

Janet could see the logic. It was the same trick that had worked in 2008-reframe the bailout as helping ordinary people rather than helping banks.

“There’s still one problem,” she said. “If we announce that the government is backstopping insurance company failures, what happens to pension funds? They’re in worse shape than the insurance companies. CalPERS is at 62% funded. Illinois is at 58%. Do we backstop them too?”

The room went quiet.

“Pension funds are different,” Thomas said carefully. “They’re state entities. The federal government can’t just take them over.”

“But if we protect insurance policyholders and don’t protect pension beneficiaries, that creates a political problem. Teachers in Illinois will ask why their pensions are getting cut while insurance policyholders in California are fully protected.”

“So what’s the answer?”

Janet pulled up a slide she’d been working on for weeks. “We need a comprehensive framework. Here’s what I propose:

Tier 1: Insurance Companies (Federal Responsibility)
– FDIC takes over failing companies
– Treasury backstops policyholder losses through Policyholder Protection Fund
– Cost: $40 billion

Tier 2: Pension Funds (State/Federal Partnership)
– States must demonstrate commitment to increasing contributions
– Federal government provides matching funds to restore solvency
– Cost: $60 billion federal, $60 billion state

Tier 3: Private Credit Funds (No Federal Support)
– These are sophisticated investors who understood the risks
– Let them liquidate and take losses
– No taxpayer money

Total federal cost: $100 billion. Spread over five years: $20 billion per year.”

Thomas was making notes. “This might work. The politics are delicate but manageable. We’re helping retirees, not helping hedge funds.”

“There’s one other component,” Janet said. “We need to make sure this never happens again. That means regulation.”

“What kind of regulation?”

“First: Cross-collateralization disclosure requirements. If you’re using the same asset to secure multiple loans, all parties must be notified. Second: Private market valuation standards. No more fund managers marking their own homework. Third: Pension fund allocation limits-no single sector can represent more than 15% of private market exposure.”

Michael from the Fed nodded. “Those are reasonable. The industry will hate them, but they’re necessary.”

After the meeting, Janet walked back to her office with Thomas.

“Do you think Congress will actually pass this?” she asked.

“If we frame it right. The key is to make it about protecting ordinary people, not about bailing out Wall Street. We emphasize the teachers, the firefighters, the middle-class families. We make the insurance companies and private equity firms the villains.”

“They kind of are the villains.”

“Sure. But we’re going to spend $100 billion of taxpayer money cleaning up their mess. That’s still a bailout, even if we don’t call it that.”

Janet thought about the last six months. About all the warnings that had been ignored. About all the risk managers who’d seen this coming and been told to wait.

“Thomas, can I ask you something? Why didn’t anyone stop this earlier?”

He stopped walking. “You mean why didn’t regulators step in when it was still manageable?”

“Yeah.”

“Because we didn’t know how bad it was until it was too late. The insurance companies didn’t disclose their full exposure. The private credit funds didn’t report accurate valuations. The pension funds delayed mark-downs. Everyone hid the problem until it couldn’t be hidden anymore.”

“So this could happen again.”

“Not with AI. That bubble’s popped. But with the next thing? Yeah. People have short memories. In five years, there’ll be some new investment theme. Everyone will pile in. Regulators will miss the warning signs. And we’ll be right back here.”

That evening, Janet sat in her Arlington apartment, eating takeout and watching CNBC. They were running a special: “The AI Crash: What Went Wrong.”

The guests were all people who’d been warning about AI for months. Sarah Chen from CalPERS. Rebecca Torres, now at the FDIC. Some professor from MIT who’d written a paper about AI economics that nobody had read.

Everyone was suddenly an expert on why the bubble had burst. Everyone had seen it coming. Everyone was surprised that others hadn’t listened to them.

But Janet knew the truth. Plenty of people had seen it coming. She’d been one of them. But seeing a crisis coming and stopping a crisis were different things.

You couldn’t stop a crisis by writing memos. You couldn’t stop it by running stress tests. You couldn’t stop it by testifying before Congress.

You could only stop a crisis by making people do uncomfortable things early, when the problem was still manageable.

And nobody wanted to do uncomfortable things. Not pension fund boards. Not insurance company executives. Not private equity partners.

Everyone wanted to wait and see if it got better.

It never got better.

Her phone buzzed. An email from Thomas Chen at Treasury. The subject line: “Framework Approved – Implementation Begins Monday.”

She opened it. The Policyholder Protection Fund had been authorized. $40 billion to protect insurance policyholders. Another $60 billion in matching funds for pension systems that committed to increased contributions.

It was happening. The intervention. The bailout that wasn’t officially a bailout.

It would stop the crisis from getting worse. It would protect millions of people from losing their retirement savings. It would restore some stability to the financial system.

But it wouldn’t change the fundamental problem.

Because in five years, or ten years, there would be another bubble. Another sector that everyone piled into. Another crisis that everyone saw coming and nobody stopped.

And Janet would be sitting in another conference room, explaining how it happened and what they should have done differently.

That was the job. Not preventing crises. Just cleaning them up.

She closed her laptop and went to bed.

Tomorrow, the real work would begin.

October 28, 2026

Sarah Chen stood in front of a Bloomberg terminal at 6:45 AM, watching the S&P 500 futures. They were down 3.2% before the market even opened.

It was the denominator effect. Finally visible. Finally undeniable.

Her phone rang. It was James Mitchell from Texas ERS.

“Sarah, we’re about to announce $12 billion in equity sales. Just wanted to give you a heads up before it hits Bloomberg.”

“How much time do I have?”

“Minutes.”

“Jesus. Okay. We’re announcing $8 billion in sales today too. I guess we’re doing this together.”

After she hung up, Sarah pulled up her spreadsheet. CalPERS had been modelling this scenario for months. The math was brutal but simple:

The Denominator Effect Explained:

CalPERS target allocation:
– 60% public equities ($180B)
– 30% private markets ($90B)
– 10% fixed income ($30B)

After private market mark-downs of $6.2B, the actual allocation became:
– 60.8% public equities ($180B)
– 28.3% private markets ($83.8B)
– 10.9% fixed income ($30B)

To rebalance to target allocations, CalPERS needed to:
– Sell $8 billion in public equities
– Use proceeds to restore private market allocation (new commitments)

But here was the problem: every major pension was doing the same calculation at the same time. And nobody wanted to make new private market commitments right now. So the $8 billion just sat in cash. Which meant the real allocation became:

– 58% equities ($172B)
– 28% private markets ($83.8B)
– 14% cash ($42B)

Which was even further from target. Which meant selling more equities. Which drove stock prices down. Which made the equity side of the portfolio worth less. Which made the percentages shift again.

It was a doom loop. And it was starting.

At 9:30 AM, the market opened. The S&P dropped 4.1% in the first hour of trading. By lunch it was down 6.2%.

Sarah watched the Bloomberg terminal with a sick feeling. This was it. This was the cascade they’d all been dreading. Not a crash driven by panic or margin calls or some external shock. Just the mechanical grinding of institutional portfolio rebalancing.

Her phone was ringing constantly now. Reporters. Board members. The Governor’s office. She ignored them all and focused on the math.

If CalPERS sold $8 billion and Texas sold $12 billion and New York sold $10 billion-and that was just the three largest pensions-they were talking about $30 billion in forced equity selling. Possibly $50 billion when smaller pensions followed suit.

In a $40 trillion equity market, that shouldn’t matter. But it mattered because everyone knew more selling was coming. And when you knew more selling was coming, you didn’t buy. You waited. Which meant prices kept falling.

By market close, the S&P was down 8.3%. The Dow was down 9.1%. The Nasdaq-heavy in tech stocks that were seen as “AI adjacent”-was down 11.4%.

$2.8 trillion in market value had evaporated in a single day.

That evening, Sarah sat in an emergency board meeting via Zoom. All eighteen board members were on the call, along with the Governor’s chief of staff.

“This is a disaster,” someone said. “We sold $8 billion in equities today and the market crashed. We’re responsible for this.”

“We’re not responsible,” Sarah said tiredly. “We’re one of dozens of institutional investors rebalancing simultaneously. This was inevitable.”

“What do we do tomorrow?” Patricia Wong asked.

“Tomorrow we sell another $2 billion to complete our rebalancing. Texas sells another $3 billion. New York sells another $4 billion. The market probably drops another 4% to 6%.”

“And then?”

“And then we’ve completed our rebalancing. But our equity portfolio is worth 15% less than it was two weeks ago. Which means our funded ratio drops from 66% to maybe 62%. Which means even deeper benefit cuts.”

The Governor’s chief of staff spoke up. “Is there any way to stop this? Some kind of coordination with other pensions to delay selling?”

“That would be market manipulation,” Sarah said. “And even if it were legal, it wouldn’t help. We have fiduciary duties to maintain target allocations. If we don’t rebalance, we’re violating our investment policy. If we do rebalance, we push markets down. There’s no good option.”

After the call ended, Sarah sat alone in her office. Outside, it was dark. The CalPERS headquarters was empty except for security guards and a few analysts working late.

She pulled up her model one more time. The one she’d been building since December. The one that showed how losses in private markets would cascade to public markets would cascade to pension funded ratios would cascade to benefit cuts.

Every step had played out exactly as she’d modeled. The only variable that had been worse than her assumptions was the Taiwan situation. That had accelerated everything by three months.

She thought about her Senate testimony. About how she’d said that CalPERS would drop from 72% funded to somewhere between 61% and 52%, depending on how bad the losses were.

They were currently at 66%, heading toward 62% after the equity market crash. If markets kept falling-and they would-they’d hit 58% or 60%.

She’d been right. Again.

It didn’t feel like a victory.

Her phone buzzed. A text from Rebecca Torres: “Pacific Life officially insolvent. FDIC taking over Friday. 2.8 million policyholders affected.”

Sarah stared at that message. 2.8 million people. Most of them didn’t follow financial markets. Most of them didn’t know what “private credit” or “cross-collateralization” meant. They just knew they’d bought life insurance or annuities from a company they thought was stable.

And now that company was being taken over by the government.

She texted back: “I’m sorry. How are you doing?”

“Unemployed as of Friday. But I’ve got three months severance and a job offer from the FDIC. They want me to help wind down the portfolio.”

“You going to take it?”

“Yeah. Someone needs to clean up this mess. Might as well be people who saw it coming.”

After they finished texting, Sarah opened her laptop and started writing. Not a memo. Not an analysis. Just thoughts.

October 28, 2026

Today the S&P dropped 8.3% because pension funds had to sell equities to rebalance their portfolios after taking private market losses. Tomorrow it will drop another 4% to 6%. By the end of the week, we’ll have lost $4 trillion in market value.

Everyone will say this was unexpected. That nobody could have predicted it. That the losses were driven by unprecedented events.

But that’s not true.

I saw this coming in December 2025. Marcus Webb saw it in December 2025. Rebecca Torres saw it in January 2026. We all saw it. We all warned people. And nobody listened.

Not because they didn’t believe us. But because acknowledging the problem would have required doing something uncomfortable. Marking down positions. Cutting allocations. Admitting losses.

So instead, everyone waited. Everyone hoped it would get better. Everyone assumed someone else would fix it.

And now we’re here. $100 billion in losses across private markets. $4 trillion in losses across public markets. Pension funds insolvent. Insurance companies failing. Retirees facing benefit cuts.

This is how financial crises happen. Not because nobody sees them coming. But because the people who see them coming aren’t in positions to stop them. And the people in positions to stop them don’t want to acknowledge the problem until it’s too late.

She saved the document. Encrypted it. Sent a copy to her personal email.

Because when this was over-when congressional hearings started and everyone asked “why didn’t anyone warn us?”-Sarah wanted there to be a record.

There had been warnings. Lots of warnings.

Nobody had wanted to listen.

October 7, 2026

Marcus Webb sat in Tom Hendricks’s office, reading the email that would end their firm. It was from their largest LP-the California State Teachers’ Retirement System.

RE: Redemption Request – Sequoia Harbor Capital Fund IV

Dear Tom,

CalSTRS is exercising its redemption rights for our $450 million commitment to Sequoia Harbor Capital Fund IV, effective immediately. As you know, our recent mark-downs in private market valuations have triggered rebalancing requirements. We must reduce our private credit exposure to meet target allocations.

We understand the fund has suspended redemptions. We believe this suspension violates the Limited Partnership Agreement Section 12.3, which permits redemptions with 90 days notice except in cases of “market emergency.” We do not believe current conditions constitute an emergency as defined in the LPA.

Our counsel will be in touch regarding next steps…

Marcus put down the email. “They’re suing us.”

“I know,” Tom said. “We’ve gotten three more of these today. Total redemption requests: $1.2 billion.”

“On a fund that’s worth maybe $1.3 billion after mark-downs.”

“Exactly. If we honor the redemptions, we have to liquidate everything. If we don’t honor them, they sue us and we liquidate everything anyway.”

Marcus pulled up the portfolio. Sequoia Harbor Capital Fund IV had started with $2.2 billion in commitments. They’d invested in 47 companies. Now:

– 18 companies had failed completely (total loss)
– 12 companies were in bankruptcy/restructuring (maybe 20 cents on the dollar)
– 9 companies were deeply distressed (maybe 40 cents on the dollar)
– 8 companies were surviving but worth less than invested capital

Only 4 companies out of 47 were actually worth more than the firm had paid for them.

“What’s the liquidation value if we sell everything today?” Tom asked.

Marcus had run these numbers a dozen times. “Maybe $800 million. But that’s if we can find buyers. In this market, we might get $600 million.”

“So LPs get back 27 cents on the dollar.”

“Maybe less.”

Tom got up and walked to the window. The view of the Bay Bridge that had once seemed so impressive now just felt like a reminder of how much money they’d spent on this office.

“How many other funds are in the same position?” Tom asked.

“All of them. Sequoia Harbor, Khosla, Founders Fund, Coatue-everyone who invested in AI. We’re all facing the same redemption requests. We’re all going to liquidate. And when we all try to sell at the same time, prices collapse further.”

“A fire sale.”

“A fire sale.”

Tom turned back from the window. “Marcus, I need to ask you something. When did you know we were finished?”

Marcus thought about it. “December. When I saw the cross-collateral structures. I knew then that when companies started failing, the losses would be worse than anyone modeled. I tried to tell you.”

“I know. I should have listened. But Marcus-even if I had listened, what would we have done? Stop investing in AI? Exit our positions? We’d have looked like fools. Everyone else was still investing. The marks were still going up.”

“We’d have looked like fools for six months. Then we’d have looked like geniuses.”

“Maybe. Or maybe we’d have been wrong and missed out on the returns. That’s the problem with bubbles. You never know you’re in one until it pops.”

Marcus’s phone buzzed. Another email. This one from New York State Common Retirement Fund. Another redemption request. $380 million.

“How much time do we have?” Tom asked.

“Our suspension lasts 180 days from June. That means December. We need to respond to all these redemption requests by December or the LPs can force liquidation through the courts.”

“So we have two months.”

“We have two months.”

After the meeting, Marcus walked through the office. Sequoia Harbor had employed 45 people when the fund launched in 2023. They’d laid off 15 in August. Now they’d have to lay off the rest.

He stopped at Karen Reyes’s desk. She was their general counsel. She’d been reviewing LPAs and redemption requests for three weeks straight.

“Karen, what are our legal options?”

She looked up from her laptop. “Honestly? Not many. The LPs have legitimate redemption rights. We suspended them citing market conditions. But several LPs are arguing that our suspension was pretextual-that we just didn’t want to give them their money back. If a judge agrees, we lose.”

“And if we fight it?”

“We spend a year in litigation and lose anyway. Meanwhile, the fund continues to decline in value, the LPs get even less money back, and we destroy what’s left of our reputation.”

“So we should just liquidate.”

“We should liquidate in an orderly way on our own terms rather than being forced to liquidate chaotically by court order.”

Marcus nodded. He’d reached the same conclusion weeks ago. But saying it out loud made it real.

“I’ll talk to Tom. We’ll announce orderly liquidation by end of month.”

That evening, Marcus sat in his expensive office with the expensive view, thinking about his career. Twenty-seven years in finance. He’d made it to partner at a $12 billion fund. He’d structured 47 deals. He’d been good at his job.

And now it was over.

His phone rang. Sarah Chen.

“Marcus, I heard about the redemption requests. I’m sorry.”

“Thanks. How’s CalPERS?”

“We’re down to 66% funded. We’ve implemented benefit restrictions. The Governor is talking about increasing state contributions but there’s no money in the budget. It’s a mess.”

“How many other pensions are in the same position?”

“Most of them. New York State Teachers is at 64%. Texas ERS is at 67%. Ohio is at 69%. Everyone’s clustered in the same range. Everyone’s facing the same problems.”

Marcus pulled up his Bloomberg terminal. He typed in a search: “Public pension funded ratios 2026.”

The results were grim:
– CalPERS: 66%
– CalSTRS: 68%
– New York State Teachers: 64%
– Texas ERS: 67%
– Ohio PERS: 69%
– Illinois TRS: 61%

Every major public pension in the country was below 70% funded. Most were below 65%.

“Sarah, what’s the total shortfall across all pensions?”

“Rough estimate? $500 billion to $700 billion in unfunded liabilities. That’s how much states and municipalities would need to contribute to get back to 100% funded.”

“They can’t afford that.”

“No. So they’ll cut benefits, raise contribution rates, and hope the markets recover.”

“And if markets don’t recover?”

“Then we’re looking at a retirement crisis. Millions of public sector workers getting 80% or 85% of their promised benefits. For the rest of their lives.”

After they hung up, Marcus sat in the darkness of his office. Outside, the city lights twinkled. Somewhere out there, hundreds of thousands of teachers and firefighters and civil servants were going to bed, not yet knowing that their retirement plans had just changed.

His phone buzzed one more time. Tom: “Board meeting tomorrow 9 AM. We’re voting to liquidate the fund. Prepare talking points for the LP call.”

Marcus stared at that message. Tomorrow they’d vote to end the fund. Next week they’d announce it to LPs. In two months, they’d liquidate everything.

And then Sequoia Harbor Capital would join the list of firms that had imploded in the AI bubble.

He opened his laptop and started drafting the talking points. But the words felt hollow:

“While we deeply regret the losses our LPs have sustained, we believe that orderly liquidation is in the best interests of all stakeholders…”

Translation: We lost all your money and now we’re shutting down.

“The unprecedented deterioration in AI sector fundamentals, combined with supply chain disruptions, exceeded all reasonable risk models…”

Translation: Nobody saw this coming, even though some people definitely saw this coming.

“Sequoia Harbor Capital remains committed to winding down the fund in a manner that maximizes value recovery for LPs…”

Translation: You’re getting back 30 cents on the dollar if you’re lucky.

He deleted it all and started over.

This time he wrote the truth:

“We failed. We invested in companies with unsustainable business models because everyone else was doing it. We ignored warnings from our own analysts because acknowledging them would have been uncomfortable. We convinced ourselves that scale would fix the economics. We were wrong. And now you’re paying the price for our mistakes. I’m sorry.”

Then he deleted that too.

Because that’s not what you said in an LP letter. You said the market moved against you. You said conditions were unprecedented. You said you did everything right but got unlucky.

You never admitted that you were part of a bubble. That you ignored the warning signs. That you put other people’s money into investments that were obviously unsustainable.

Even when everyone already knew.

September 3, 2026

David Huang accepted the Meta offer on a Tuesday. By Friday, he was already doubting the decision.

The offer was good-$380,000 base plus $200,000 in RSUs. More than he’d ever made in his life. Enough to finally propose to Lisa without worrying about money. Enough to stop checking his bank account every morning.

But something was bothering him.

He was sitting in his kitchen on Saturday morning, scrolling through his LinkedIn feed, when he saw it:

“GPU Spot Prices Reach $180,000 Per H100”

He blinked. Read it again. Opened the article.

NVIDIA H100 GPUs, which sold for $40,000 each eighteen months ago and were trading at $50,000 in July, have spiked to $180,000 on secondary markets. The increase reflects severe supply constraints following shipping delays through the Taiwan Strait and hoarding by major tech companies.

“Everyone’s trying to lock in supply before it gets worse,” said one trader who spoke on condition of anonymity. “Companies that need GPUs for Q4 are paying whatever it takes. Companies that don’t have cash are out of luck.”

David did the math. $180,000 per GPU. Artemis had needed 200 GPUs to train their models. That would now cost $36 million just for the hardware. Their entire Series B had been $35 million.

He called Jamie.

“You seeing these GPU prices?” David asked.

“Yeah. It’s insane. There are companies that were barely surviving before-they’re dead now. Can’t afford new equipment. Can’t expand. Can’t even maintain their existing infrastructure because replacement costs are through the roof.”

“How many companies fail because of this?”

“I’ve been tracking about thirty companies in our network. Half of them needed to buy equipment in Q3 or Q4. At $180K per GPU? They can’t. Maybe ten companies survive. The rest are done.”

After they hung up, David pulled up a spreadsheet. He still tracked the AI sector even though he wasn’t working in it anymore. Old habits.

He listed out companies he knew were in trouble:

– Anthropic: Still burning billions, but they’d pre-ordered GPUs at old prices. Probably survive.
– Inflection: Already bankrupt.
– Character.AI: Struggling but might get acquired.
– Cohere: Rumoured to be in acquisition talks with Google.
– Adept: Just laid off 40% of staff. Probably fails.
– Hugging Face: Mostly open source, not hardware dependent. Survives.
– Runway: Content generation, might survive if they can raise.

He kept going. By the time he finished, his spreadsheet showed:
– 15 companies likely to survive
– 23 companies likely to fail
– 12 companies might survive if they get acquired

That was 35 companies failing or getting acquired at fire-sale prices out of 50 companies. A 70% failure rate.

He thought about Sarah Chen’s Senate testimony. She’d predicted 50% to 70% failure rates. Everyone had said she was too pessimistic.

She’d been right.

His phone buzzed. Rebecca Torres.

“David, I need your help with something. Can we talk?”

They met at a coffee shop in Mountain View. Rebecca looked exhausted. She’d lost weight. Her Pacific Life badge was clipped to her belt but it looked like she was wearing it out of habit rather than pride.

“How are you?” David asked.

“I’m about to lose my job. Pacific Life is going to be taken over by the FDIC sometime in October. But that’s not why I called. I need you to explain something to me.”

She pulled out her laptop and showed him a spreadsheet. “I’m trying to model the impact of GPU price spikes on our portfolio companies. But I don’t understand the technical side well enough. Walk me through what happens when a company can’t afford new GPUs.”

David looked at her screen. She had a list of 30 companies with estimates of their GPU needs and cash positions.

“Okay,” he said. “There are basically three scenarios:

“Scenario One: Company has enough GPUs and enough cash to maintain them. They survive but can’t grow. That’s maybe 20% of companies.

“Scenario Two: Company needs to expand their GPU fleet to serve more customers or train bigger models. They can’t afford new GPUs at $180K each. So they stop growing. Their existing customers are fine but they can’t take on new customers. Eventually, they get acquired or they shrink into irrelevance.

“Scenario Three: Company’s existing GPUs are leased, not owned. The lease comes up for renewal. The lessor wants to re-price at current market rates. The company can’t afford it. They lose access to their infrastructure. They shut down.”

“Which scenario is most common?”

“Scenario Three. Most companies leased their GPUs because buying them outright required too much capital. Now they’re trapped. They’re making payments on equipment that’s worth four times what they agreed to pay. The lessors know this. Some of them will renegotiate. Most won’t.”

Rebecca was taking notes. “How fast does this play out?”

“Most leases are two or three years. A lot of companies did their leases in 2024. That means renewals in 2026-2027. We’re in the window now. I’d guess twenty companies fail in Q3-Q4 just from lease renewals they can’t afford.”

“Jesus.”

They sat in silence for a moment, drinking bad coffee.

“David, can I ask you something? Do you think AI was ever real? Or was it all just hype?”

David thought about Artemis. About the lung cancer their software had detected. About the lives they’d probably saved.

“The technology is real. The medical imaging AI we built actually worked. It was better than human doctors. But the business model was broken. We lost money on every customer. And that was true when GPUs cost $40,000. Now that they cost $180,000? It’s impossible.”

“So the technology works but the economics don’t.”

“Exactly. And everyone-VCs, pension funds, insurance companies-bet on the idea that scale would fix the economics. That if we just grew fast enough, we’d eventually make money. But scale doesn’t fix unit economics when your costs grow faster than your revenue.”

Rebecca closed her laptop. “I spent twelve years as a risk manager. I’ve seen a lot of bad investments. But I’ve never seen anything quite like this. Everyone knew the economics didn’t work. Everyone kept investing anyway.”

“Because everyone else was investing. Classic bubble psychology.”

“And now we’re all paying for it.”

After they left the coffee shop, David walked to his car. He sat in the driver’s seat for a moment, thinking about Meta. About the offer he’d accepted. About the fact that he was joining a company that had spent $10 billion on AI infrastructure and had almost nothing to show for it.

Meta would be fine. They had their social media business. But their AI investment? Probably a write-off. Just like everyone else’s.

He started the car and drove home. On the radio, NPR was running a story about CalPERS announcing benefit cuts for retirees. The reporter was interviewing a retired teacher who’d worked for thirty-five years and was now facing a 15% reduction in her pension.

David turned off the radio.

He didn’t need to hear it. He already knew how this ended.

August 15, 2026

Sarah Chen had been dreading this day for six months. Now it was here.

She stood in front of the CalPERS board, presenting the Q2 portfolio review. The room was packed. All eighteen board members present. The press was waiting outside. The Governor’s office had sent a representative.

Everyone knew this was going to be bad.

“As of June 30th, 2026,” Sarah began, “CalPERS’s funded ratio stands at 66.3%, down from 72% at year-end 2025.”

She let that number sit for a moment. Under California law, dropping below 70% triggered benefit restrictions. At 66%, they were in crisis territory.

“The primary driver is mark-downs in our private market portfolio. Specifically, AI-related investments across private equity and private credit funds. Total losses: $6.2 billion.”

Board member Patricia Wong raised her hand immediately. “Sarah, in your testimony before the Senate in February, you estimated losses of $4 to $6 billion. This is at the high end. What changed?”

“Two things. First, more companies failed than our base case assumed. We projected 30% failure rate. Actual: 45%. Second, the Taiwan Strait situation disrupted supply chains faster than we modelled. Companies that we thought would survive Q3 are failing now because they can’t get equipment.”

James Rodriguez, the gubernatorial appointee who’d questioned her in April, leaned forward. “In April, we voted to delay mark-downs until June 30th. If we’d marked down then, would the losses have been smaller?”

Sarah had prepared for this question. She had the analysis ready. But saying it out loud would be admitting that the board’s decision to wait had cost CalPERS money.

“If we’d marked down in April, using the data available then, our estimated losses would have been $5.1 billion. The additional $1.1 billion in losses reflects further deterioration from April to June.”

“So waiting cost us a billion dollars.”

“Waiting cost us transparency. The billion dollars was already lost. We just hadn’t admitted it yet.”

The room was quiet.

Martin Zhao, the Chief Investment Officer, jumped in. “Sarah, let’s be clear-these mark-downs reflect current market conditions. If the AI sector stabilizes, these valuations could recover-”

“They won’t recover,” Sarah said flatly. She was done being diplomatic. “The companies are failing. The collateral is worthless. The cross-collateral disputes are tying up assets for years. We are not getting this money back.”

She pulled up her next slide. “Here’s what 66% funded means for CalPERS members. Under California Government Code Section 20221, we must implement benefit restrictions:

– New retirees: Benefits calculated at 85% of full formula
– Current retirees: No COLA adjustments for three years
– Active members: Contribution rates increase 2.5%
– State contribution: Increases $6.2 billion annually

“The political implications are significant. Teachers, firefighters, civil servants-they all take cuts. The state budget takes a $6 billion hit. And CalPERS is not unique. Every major pension fund has similar exposure.”

Board member Michael Torres from the firefighters’ union stood up. “Sarah, you’re telling me that firefighters who worked thirty years are going to retire with 15% less than they were promised?”

“Yes.”

“Because CalPERS invested in AI companies that failed?”

“Because CalPERS, like every major institutional investor, allocated to private markets to chase returns. And private markets allocated to AI because that’s where the returns were supposed to be. And the AI companies failed because their business models didn’t work.”

“Who authorized these investments?”

Sarah looked at Martin Zhao, who was studying his notes very intently.

“The board approved the overall allocation to private markets-30% of assets. Investment staff, guided by the CIO, selected specific funds and managers. Those managers selected portfolio companies. At no point did this board vote to ‘invest in AI.’ You voted to invest in private markets that generate 7% returns. The fund managers decided AI was how to get those returns.”

Patricia Wong spoke up again. “Sarah, in February you warned us this was coming. You recommended marking down immediately. We voted to wait. That was a mistake. I want it on the record that you were right and we were wrong.”

“Thank you, but that doesn’t help our members.”

“No. But it matters for the next crisis. Sarah, what do we do now?”

Sarah pulled up her final slide. It was titled “Path Forward” but it might as well have been titled “Damage Control.”

“Three immediate actions:

“One: We announce the 66% funded status today, along with the benefit restrictions required by law. We don’t hide it. We don’t spin it. We tell the truth.

“Two: We implement an immediate freeze on all new private market commitments. No new funds, no new investments, until we’ve fully assessed remaining exposure.

“Three: We coordinate with other large pensions-New York, Texas, Ohio-to mark down simultaneously. If we all do it together, we avoid the perception that CalPERS is uniquely troubled.”

“Have you talked to the other pensions?” someone asked.

“I’ve talked to Sarah Chen at New York State Teachers’ Retirement System-different Sarah-and James Mitchell at Texas ERS. They’re both planning to mark down in their Q3 reporting. We’re all seeing the same losses.”

After the board meeting, Sarah stood outside the CalPERS headquarters, facing cameras for the first time since her Senate testimony.

“Today, CalPERS’s board voted to acknowledge the reality of losses in our private market portfolio. Our funded ratio is 66.3%. This requires us to implement benefit restrictions under California law. We deeply regret the impact this will have on our members, but we have a duty to be transparent about our financial condition.”

A reporter shouted: “Ms. Chen, you warned about this in February. Why didn’t CalPERS act sooner?”

“CalPERS acted consistent with standard industry practice-marking positions based on fund manager reports at quarter-end. The speed of deterioration in the AI sector exceeded most projections.”

“Do you expect other pensions to face similar problems?”

“Every large institutional investor has exposure to private markets. Many will face similar challenges. This is not a CalPERS-specific issue. It’s a sector-wide repricing of private market valuations.”

“What do you say to teachers and firefighters who will receive reduced benefits?”

Sarah thought about her prepared talking points. About how she was supposed to express sympathy but remain optimistic. About how she was supposed to say that CalPERS remained well-positioned for the long term.

Then she thought about the firefighter who’d stood up in the board meeting. About the teachers who’d worked thirty years. About the promises that were being broken.

“I say that we failed you. We chased returns in private markets because we couldn’t get them anywhere else. We invested in companies with unsustainable business models because everyone else was doing it. We ignored warning signs because acknowledging them would have been uncomfortable. And now you’re paying the price for our mistakes. I’m sorry.”

That evening, her phone exploded. Texts from pension fund managers across the country. Emails from reporters. Voice mails from fund managers who wanted to “provide context” on the mark-downs.

But one text stood out. From Marcus Webb: “You told the truth. Again. Thank you.”

She texted back: “How many other pensions are marking down?”

“All of them. New York announced today. Texas tomorrow. Within a week, every major pension in the country will have admitted AI losses. Total: $40 to $50 billion.”

Sarah stared at that number. $40 to $50 billion in admitted losses. And that was just pensions. Insurance companies would be another $20 billion. Private equity funds, another $30 billion.

They were talking about $100 billion in losses. Across the entire ecosystem. All because everyone had made the same bet at the same time.

And the Taiwan situation meant it was about to get worse.

July 22, 2026

Janet Rodriguez had been reviewing insurance company filings for nineteen years. She’d developed what her colleagues called a “sixth sense” for companies in trouble. It wasn’t magic-it was pattern recognition. You learned to spot the warning signs. The vague language. The sudden changes in investment strategy. The reserves that didn’t quite add up.

She was seeing all those warning signs now. In twelve different companies.

Her team had worked seventy-hour weeks for two weeks straight. Now they sat in a conference room in FDIC headquarters, presenting their findings to the Deputy Director and three senior officials from Treasury and the Federal Reserve.

“Let’s start with Pacific Life,” Janet said, pulling up the first slide.

“Capital position as of Q2: $7.25 billion in surplus. That’s above the regulatory minimum of $6.8 billion, but barely. Their AI-related exposure is $4.2 billion across private credit funds. Based on current default rates and the Taiwan supply disruption, we estimate they’ll take $2.1 billion in losses over the next six months.”

“That puts them at $5.1 billion,” someone said. “Below the minimum.”

“Correct. Pacific Life becomes insolvent by Q4 unless they raise capital or dramatically reduce exposure. Neither seems likely given market conditions.”

She clicked to the next slide. “MetLife. Similar story. $18 billion in surplus, $9.2 billion in AI exposure. Estimated losses: $4.6 billion. They’ll stay solvent but will need to file a remediation plan.”

She went through all twelve companies. By the time she finished, the picture was clear:

– 3 companies would be insolvent by Q4
– 5 companies would need remediation plans
– 4 companies would remain adequately capitalized but stressed
– Total estimated losses across all twelve: $21 billion

The Deputy Director, Sarah Morrison, sat back in her chair. “Twenty-one billion in losses. How much of that flows through to policyholders?”

“If we take over the three insolvent companies, we can protect most policyholders through the guaranty fund system. But there will be benefit reductions. Life insurance death benefits might be paid at 90 cents on the dollar. Annuities might face payment delays.”

“How many policyholders are we talking about?”

Janet had been dreading this question. “Across the three companies: 6.2 million policyholders.”

The room went quiet.

Thomas Chen, the Treasury Undersecretary who’d been listening silently, spoke up. “Walk me through the contagion scenario. These three fail. Then what?”

Janet pulled up her cascade model. “When we take over Pacific Life, MetLife, and Prudential will have to mark down similar positions. That triggers further losses. They stay solvent but they stop writing new business, which means market share shifts to smaller, weaker companies. Meanwhile, pension funds see insurance companies failing and start questioning their own private market valuations. CalPERS, New York State, Texas-they all have to mark down. That triggers the denominator effect-”

“Explain the denominator effect,” Thomas interrupted.

“Pension funds have target allocations. For example, 60% public equities, 30% private markets, 10% bonds. When private market values drop, the percentages shift. Suddenly you’re 50% public, 40% private, 10% bonds. To rebalance, you have to sell public equities. That pushes down stock prices. Which hurts the public equity side of your portfolio. Which makes the funded ratio worse. It’s a doom loop.”

Thomas was making notes. “How much would pensions have to sell?”

“We estimate $50 to $80 billion in forced equity sales over six months if the three insurers fail.”

“Impact on public markets?”

“S&P down 8% to 12%. Maybe more if it triggers margin calls and systematic selling.”

The Fed official, Michael Park, leaned forward. “Janet, you’re describing a cascading financial crisis triggered by the failure of three insurance companies. What’s your confidence level in this scenario?”

Janet thought about the last two weeks. About the models her team had built. About the sleepless nights running scenarios. About Pacific Life’s risk assessment that showed them heading toward insolvency.

“High confidence on the first-order effects-the three companies failing. Medium confidence on the second-order effects-pension marks and equity sales. Low confidence on whether it becomes truly systemic. That depends on policy response.”

Sarah Morrison, the Deputy Director, closed her laptop. “Okay. Here’s what we’re going to do. Janet, I need you to reach out to Pacific Life, MetLife, and Prudential quietly. Tell them we’re conducting enhanced supervision reviews. Get current data on their positions. See if they’re taking any steps to raise capital or reduce exposure. Don’t tell them we think they’re insolvent-just that we’re monitoring closely.”

“And if they are insolvent?”

“Then we move to resolution planning. But quietly. The last thing we need is public panic about insurance companies failing.”

Thomas from Treasury spoke up. “I need to brief the Secretary. What’s our timeline?”

“Pacific Life probably has 90 days before they breach capital requirements. The others, 120 to 150 days. So we’re looking at October, November timeframe for the failures to become public.”

“Unless something accelerates it.”

“Right. Unless something accelerates it.”

After the meeting, Janet stood in the hallway with Michael Park from the Fed.

“You really think it’s going to cascade?” he asked.

“I’ve been doing this for nineteen years. I’ve seen a lot of problems. But I’ve never seen a situation where every major institution made the same bet at the same time. Usually risk is distributed. This time, it’s concentrated.”

“What happens if we don’t intervene?”

Janet thought about her models. About the spreadsheets showing capital ratios dropping and reserves depleting and benefit reductions cascading through the system.

“If we don’t intervene? Six million policyholders take losses. Pension funds drop to 60% funded or lower. State and local governments face budget crises. Retirees get 80% of their promised benefits. The political fallout makes 2008 look manageable.”

“And if we do intervene?”

“Then we need to figure out how to bail out insurance companies without calling it a bailout. Because that’s going to be a very hard sell politically.”

Michael smiled grimly. “That’s why they pay us the big bucks.”

But as he walked away, Janet thought about her salary. She made $160,000 a year. Not exactly big bucks for someone trying to prevent a multi-trillion dollar crisis.

She went back to her office and started drafting the outreach emails to Pacific Life, MetLife, and Prudential. She kept the language careful and neutral:

“As part of our ongoing supervisory responsibilities, the FDIC is conducting enhanced risk reviews of insurance companies with significant private market exposure…”

She knew what they’d think when they received it. They’d think: Oh shit. The regulators know.

And they’d be right.