Dennis Hendrickson
AI & Tech · October 13, 2026

Hock Tan Calls AI Investors Gamblers While Broadcom Bankrolls the Casino

Broadcom's CEO warned that AI financing, not chips, could be the next bottleneck, and the fallout is already hitting credit markets, small business lending and retirement accounts.

On Monday, Broadcom CEO Hock Tan stood up at a DBS conference in Singapore and said something CEOs almost never say out loud: that a lot of the money pouring into AI is dumb money. "A lot of investors have no clue what they are investing in," he said, comparing the AI trade to a casino. He also said the industry could hit a wall not because chips run out, but because capital does. Intel's Lip-Bu Tan, sitting next to him, said hardware will stay scarce for two to three more years. Put those two statements together and you get the chips will be hard to get, and the money to pay for them will be even harder.

Markets did not shrug this off. The SOXX chip index fell 2.6% and JPMorgan's AI memory basket dropped 4.3%, on a day when more than 60% of the S&P 500 actually went up. That is not a broad market worry. That is a worry about one specific corner of the economy: AI chips and the debt financing them.

What a Chip Sell-Off Means for Tech Hiring

If you work in semiconductors, data centers, or anywhere in the AI supply chain, this matters to your job security even if you never look at a stock chart. Companies do not keep hiring and building at the same pace once their stock and their credit get this shaky. Broadcom's bonds hit all-time lows after its financing deals became public, and its five-year credit default swap, which is basically the price of insuring against Broadcom defaulting, printed a record 136 basis points last week. Oracle's hit a record 261 basis points. Nearly 500 billion dollars of 2026 AI borrowing is getting repriced right now.

When lenders start charging more to insure a company's debt, that company gets more careful with cash. Careful with cash means slower hiring, more contractors instead of full-time roles, and projects that get paused pending financing. If you are job hunting in this space, do not assume the hiring slowdown talk from earlier this year is over. Watch the credit spreads, not just the press releases.

Why Tighter AI Credit Squeezes Everyone Else's Loans

Here is the part that should worry small business owners who have nothing to do with AI. Goldman's Rich Privorotsky said last week that the AI borrowing binge is now crowding out everyone else in the capital markets. Credit markets are not an endless pool of money. When trillions of dollars get committed to AI chip financing, SPVs and data center buildouts, the lenders and underwriters doing that work have less appetite and less balance sheet left over for financing a contractor's new truck, a retailer's inventory line, or a franchise buildout.

We already saw one crack. Firmus, a Nvidia-backed data center operator in Australia, first cut its IPO price by 25%, then pulled the listing entirely and said it would raise money in private markets instead.

"Bring on the Apollo life insurance policies" was the joke, but it points at something real: when public markets will not fund a deal, it moves into private credit, often backed by insurance company balance sheets. That is a slower, more expensive, less transparent way to raise money, and it is becoming the default for anyone competing with AI for capital.

Where This Looks Exactly Like a Bad Underwriting File

I run an insurance agency, so SPVs backed by insurance money get my attention fast. Insurance companies take in premium and policyholder money and are supposed to invest it conservatively, because that money has to be there decades later to pay claims and policy benefits. When that capital gets routed into chip-collateralized structures for money-losing AI labs, you are stacking long-duration, supposedly safe liabilities on top of some of the riskiest financing in the market.

Broadcom alone is now behind roughly 145 billion dollars in chip debt: a financing package of more than 50 billion for OpenAI, 60 billion syndicated for Anthropic, and a 35 billion dollar platform called Big Sky, built with Apollo and Blackstone.

In my world, when one underwriter is both the biggest seller to a customer and the biggest lender to that same customer, that is a conflict you flag on the file, not something you brag about on a conference panel. Hock Tan calling investors gamblers while his own company arranges the chips on credit for its own customers is exactly that kind of conflict.

What This Does to Everyday 401(k)s and Retirement Accounts

You do not need to own Broadcom stock directly to be exposed to this. Goldman's derivatives desk says the Magnificent Seven tech stocks now make up about 22% of total U.S. prime brokerage exposure, the most since early 2022, and semiconductor exposure has doubled to about 12% from 6% at the start of the year. Roughly 115 billion dollars sits in leveraged semiconductor ETFs, and JPMorgan says sentiment on chip stocks is above the 99th percentile, meaning almost nobody is this bullish this often.

That concentration lives inside index funds, target date retirement funds and 401(k)s that millions of people assume are diversified. When a sector this concentrated and this debt funded wobbles, it does not stay contained to people who day trade chip stocks. It shows up in retirement account balances for people who have never heard of an SPV.

What to Actually Watch From Here

None of this means AI is fake or that chip demand disappears. It means the financing underneath it is stretched thinner than the stock charts suggest, and that stress shows up first in credit markets, then in hiring, then in your account statement.

Sources