The Numbers Are Getting Embarrassing

Alphabet just ran the largest equity offering in history to fund AI infrastructure it admits it can't build fast enough. Broadcom told Wall Street AI revenue would triple — and got punished for it. At some point, you have to ask what game we're actually playing.

Markets · June 5, 2026

---

Hock Tan stood up on Tuesday night and told the world that Broadcom's AI revenue would triple in a single quarter — from $10.8 billion to $16 billion — and the stock fell 14% by Thursday morning. Not because the number was small. Because someone, somewhere, had written down $17.2 billion, and the difference between what happened and what was imagined was enough to wipe roughly $110 billion in market value off a company that grew AI sales 143% year-over-year.

That's the market we're in. The only direction left is vertical, and anything less than vertical reads as a warning sign.

AVGO drop	–14% · After 143% YoY AI revenue growth
GOOG equity raise	$84.75B · Largest equity offering in history
Hyperscaler capex '26	~$800B · Goldman Sachs estimate

The same 48 hours handed us the other bookend: Alphabet pricing an $84.75 billion equity raise — upsized from the originally announced $80 billion because demand exceeded expectations even for that number — making it the single largest equity capital transaction ever executed. Bigger than Petrobras in 2010. Bigger than anything. The stated purpose is scaling AI compute infrastructure to meet "unprecedented customer demand." The company says supply is already failing to keep up. In the same breath, it's projecting capex of $180 billion to $190 billion this year, rising further in 2027. And per PIMCO, hyperscaler AI spending in 2026 will consume roughly 94% of their collective operating cash flow.

"Under-investing is an existential risk; over-investing is merely expensive." That line, from an equity capital markets banker quoted in coverage of the Alphabet deal, is doing a lot of work. It's also the logic that every bubble in history has used to justify itself.

The Berkshire Hathaway angle deserves a moment. Warren Buffett's firm — now run post-Buffett, though carrying the institutional weight of his name — agreed to buy $10 billion of Alphabet stock at a 6.5% discount to market. Berkshire had been building a position since Q3 2025 and is now one of Alphabet's largest shareholders. The old Berkshire would have called this speculation. The new Berkshire is writing nine-figure checks into the AI infrastructure buildout alongside Goldman, JPMorgan, and Morgan Stanley. The symbolism of that shift shouldn't be dismissed. When the last fortress of value investing joins the arms race, the arms race has become consensus.

Zoom out and the picture gets harder to ignore. The Dow tagged a new all-time high at 51,657 on Wednesday. The S&P 500 pushed its Q2 gain past 16%. S&P 500 companies just reported Q1 earnings growth of 28.6% — the highest profit margin on record since FactSet started tracking in 2009. On the surface, this looks like vindication. Underneath it, the Technology sector did essentially all the work in May, jumping nearly 20% while eight of the eleven S&P 500 sectors actually declined. Energy fell 5.6%. Utilities dropped 5.2%. Financials were negative. This is not broad-based prosperity. It's a vertical column of capital stacked on a narrow base of names, with the rest of the economy somewhere between treading water and slowly sinking.

And then there's SpaceX, which expects to price its IPO next week at a valuation targeting the neighborhood of $2 trillion. The company lost $4.3 billion in the first quarter of 2026 alone. It trades, on a price-to-sales basis, at approximately 94 times revenue — a multiple that makes Palantir look conservative. The pitch to investors is that SpaceX has identified, in its own words, "the largest actionable total addressable market in human history," a figure it puts at $28.5 trillion. Most of that is AI enterprise applications. So yes, SpaceX is now also an AI company. The rocket business is incidental.

Every loss-making company with a satellite network and a data center division is an AI company now. The question nobody is asking loudly enough is: who is the actual customer, and what are they actually buying?

The Broadcom selloff is instructive precisely because it happened despite genuine numbers. AI revenue more than doubled. EPS beat consensus. Q3 guidance calls for AI revenue to triple. Hock Tan confirmed six hyperscale customers including Alphabet, Meta, and OpenAI — the same entities currently spending themselves into a capex singularity. The punishment wasn't for bad results. It was for failing to extend the narrative far enough into the future. The market has become an engine that prices tomorrow's expectations, not today's reality — and tomorrow's expectations keep compounding faster than any business can physically build.

Meanwhile: job cuts are rising. U.S. employers announced 97,000 layoffs in May 2026, the highest for that month since 2020, running for a third straight month of increases. Consumer sentiment sits at record lows per the University of Michigan. The manufacturing PMI shows a falling order book. The savings rate is being squeezed. None of this appears in the equity indices, because the equity indices have, for all practical purposes, become a poll on one question: do you believe the AI buildout justifies its cost?

Right now, the answer is yes — with the small footnote that if Broadcom guides for AI revenue to triple and misses an analyst's spreadsheet by $1.2 billion, you lose 14% of the company in a session. The confidence is total, until it isn't. The bar keeps rising because the narrative keeps demanding it. And the narrative keeps demanding it because $800 billion in annual capex requires a story large enough to justify writing the check.

The check is being written. The story is being told. The only thing missing is the proof that the infrastructure being built will generate returns commensurate with the capital being destroyed in building it. We may not have to wait very long to find out. Or we may be having this exact conversation, with larger numbers, for years.

---

Data: CNBC, SEC filings, FactSet, Goldman Sachs, PIMCO estimates, Tickeron, 247WallSt. All figures current as of June 4–5, 2026.