Are AI Companies Too Big to Fail?
The term “Too Big to Fail” (TBTF) became popular after the 2008 financial crisis, when the government bailed out failing banks. These institutions were so deeply woven into the economy that their collapse would trigger instability.
Today’s AI giants (Google, OpenAI, NVIDIA, Meta, etc.) aren’t banks. But they’ve become critical to both market performance and economic growth. When I see the web of investments and partnerships connecting these companies, I can’t help but wonder: if something breaks, will they get bailed out?
Here are the top criteria that would determine whether something is TBTF:
Size: How large is the company relative to the overall economy?
Interconnectedness: How dependent are other entities on this company? How intricate are its operations and relationships?
Critical function: Does the company provide essential services society can’t function without?
Former Fed Reserve Chair Ben Bernanke put it this way: “A too-big-to-fail firm is one whose size, complexity, interconnectedness, and critical functions are such that, should the firm go unexpectedly into liquidation, the rest of the financial system and the economy would face severe adverse consequences.”
Today, I’m going to apply this TBTF framework to AI companies.
(And to be clear: when I say “AI companies,“ I’m talking about the infrastructure and foundational model providers like the big tech hyperscalers and OpenAI/Anthropic, not the application layer startups that build on top of these platforms).
A) Size: AI Drives Our Retirement Accounts
Tech companies now dominate the S&P 500, where most pension funds and 401(k)s invest. As you can see in the chart below, the top six companies by market cap globally are all tech, with their AI strategies : NVIDIA, Microsoft, Apple, Alphabet, Amazon, and Meta. This concentration extends beyond the obvious AI names. Even an oil giant like Saudi Aramco is looking to diversify away from oil to AI investments, and chip designers like Broadcom and manufacturers like TSMC are effectively AI infrastructure plays.
This isn’t like 2008, where our money sat directly in failing banks or our mortgage was owned by them. But if these tech companies tank, retirement accounts suffer too.
The scale of this exposure is massive. CalPERS, America’s largest public pension fund with $157 billion in equity holdings, has the Vanguard S&P 500 ETF as its single largest holding at 9.4% of its equity portfolio. Major index funds like Vanguard’s Total Stock Market Index Fund hold over $2 trillion in assets, heavily weighted toward the same tech giants.
Apollo, a major asset manager, recently published a report on “the extreme weight of AI in the S&P 500”. The concentration is historically high. The top 10 companies, most AI-focused, are driving 55% of market cap gains since 2021. The “Magnificent 7” tech stocks have massively outperformed the other 493 S&P companies. A large part of this rally is the AI story.
The impact goes beyond stock markets to GDP growth itself. Harvard economist Jason Furman estimates that without AI, US GDP growth would have been 0.1% in the first half of 2025. Morgan Stanley Wealth Management’s chief investment officer Lisa Shalett estimates that data center-linked spending is adding roughly 100 basis points to U.S. real GDP growth, with hyperscaler capex on data centers rising fourfold in recent years to nearly $400 billion annually.
When a handful of companies drive both the stock market and economic growth, their failure becomes everyone’s problem.
B) Interconnectedness: The AI Money Circle
In addition to the sheer size, these companies are deeply intertwined.
Recent examples: NVIDIA investing up to $100B in OpenAI. OpenAI signing a $300B cloud deal with Oracle. Oracle spending billions on NVIDIA chips. NVIDIA is making minority investments across the AI ecosystem.
And whenever there’s an AI-related announcement, stock markets react immediately. For example, on October 6 2025, AMD stock surged over 30% after announcing its partnership with OpenAI, adding roughly $80B in market cap.
The chart above shows the circular dependency. These companies are customers, investors, and suppliers to each other. If one major player stumbles in this circular machine, the shockwaves ripple through the entire network.
This creates a different kind of complexity than 2008. AI companies don’t have opaque financial derivatives, but training frontier models requires coordination across data centers, energy grids, and chip supply chains that can’t be quickly replaced.
C) Critical Function: Not Yet, But Getting There
This is where AI diverges from 2008’s banks. Financial institutions were critical infrastructure. Without them, the economy stops.
AI isn’t there yet. While Fortune 500 adoption is nearly universal, a recent MIT study shows 95%+ of AI pilots fail to scale beyond testing. Companies and consumers are still figuring it out. AI feels more “nice to have” than “can’t live without”, at least today.
But AI has become a national priority whether it’s critical to our daily lives or not. The US and China are in an AI arms race, with the AI Action Plan from July laying out how strategically important this is becoming. The EU, UAE, and other regions are pouring billions into AI, making it central to their economic strategies.
What’s more telling is that the US government is already treating these companies as critical infrastructure through its own dependence on them. The Department of Defense (DoD) has partnered with Anthropic, Google, OpenAI, and xAI to accelerate DoD AI adoption for national security. When the Pentagon builds its AI strategy around private companies, those companies become essential.
My Concluding Thoughts
AI companies are trending toward TBTF status. They’re massive, interconnected, and becoming strategically essential. As we grow more dependent on AI and as these companies continue driving market performance and economic growth, the case strengthens.
If something breaks (and I hope it doesn’t), I can see the argument for government intervention. The key question then becomes: at what price?