Nvidia's $150 Billion Buyback: Record Cash for Shareholders, Questions for AI

Nvidia authorized a record $150 billion stock buyback on Monday, lifting its total repurchase capacity to $235 billion through fiscal year 2028 — the largest such authorization in corporate history. The chipmaker reported $96.2 billion in revenue for the second quarter of fiscal 2027, up 106% year over year. Those two numbers frame the central tension of the AI boom: one company is collecting extraordinary cash while its customers borrow heavily to pay it.

Nvidia dominates the supply of AI-computing infrastructure — the GPUs (graphics processing units, the chips that train and run AI models), servers and software that power AI data centers. It is, by the source's account, the only company in the AI ecosystem currently making real monetary gains. During the quarter, Nvidia returned roughly $26 billion to shareholders through repurchases and dividends, and CEO Jensen Huang now ranks among the world's ten richest people.

Why does this matter beyond Nvidia's shareholder base? Because the buyback lands at a moment when the financial foundation of the AI buildout is showing strain, and the record authorization functions as a public signal that Nvidia's board expects the cash to keep flowing.

Who is actually making money in AI?

Not the startups. The AI labs and startups buying Nvidia's compute still cannot demonstrate to investors that they can generate standalone revenue, according to the source. The biggest spenders outside Nvidia — the so-called hyperscalers (the large cloud providers building out AI capacity) — face hundreds of billions of dollars in new debt, with around $300 billion of direct bond issuance expected in 2026 alone.

A tracker called Is AI Profitable Yet?, built by a single developer, aggregates cumulative monthly spending versus revenue across most major AI companies. Its question is blunt: "Has the AI industry, as a whole, made back the money it's poured into AI so far?" The answer, as of the reporting, is no.

"AI startups are burning through investor cash and handing it straight to Nvidia to buy chips, even though the startups themselves aren't making a profit," said Robin Wigglesworth, a Financial Times reporter and author of the upcoming book A Fabulous Debt, a history of the bond market.

Wigglesworth also flagged the circularity inside the ecosystem. Circularity, in this context, means capital recycling among a small group of companies that serve simultaneously as one another's investors, customers, suppliers, lenders and backbacks. Nvidia sits at the top of that loop with the most advantaged position in the supply chain: it sells the picks and shovels of the AI gold rush, and gets paid whenever any platform, startup or enterprise builds or upgrades AI capacity — before anyone proves the model can cover its costs.

The takeaway in plain words: Nvidia profits from construction, not from whether anything built actually works.

What does a $150 billion buyback actually commit Nvidia to?

Less than the headline suggests. Ed Zitron, an AI critic and host of the podcast Better Offline, called the move "an attempt to calm very nervous investors around AI." He noted that a buyback authorization is not a legal agreement — it commits Nvidia to nothing. The company is not obligated to buy that many shares, or any shares at all.

That caveat is not hypothetical. Nvidia has made large promises before, including a $100 billion data-center deal with OpenAI that never materialized, according to the source. An authorization this size is a signal, not a purchase order, and readers should treat the $150 billion figure accordingly.

The board's message, though, is clear: management expects cash flows to remain robust even as warnings about an AI bubble and business-model threats multiply. In effect, the buyback is a confidence operation aimed at Wall Street.

The takeaway: the record number is designed to reassure markets, but the fine print obligates Nvidia to almost nothing.

Why launch an AI security platform the same day?

On Monday, alongside the buyback, Nvidia introduced a new security platform designed to stop autonomous AI agents (software systems that act on their own) from behaving unpredictably or going rogue. The company described the open-source system as having guardrails against past security failures.

Paris Marx, tech critic and host of the podcast Tech Won't Save Us, argues the timing is strategic. The platform, Marx said, is designed "to take advantage of the current discourse around AI threats to get some positive headlines" — even if it never becomes an important part of the industry or a revenue driver. By addressing doomsday scenarios now entering mainstream discussion, Nvidia protects its own boom and helps ensure market demand does not collapse.

Wigglesworth put it similarly: "Launching something to ameliorate the dangers of rogue AI — or at least being seen to be ameliorating that risk — therefore makes sense, to keep the show on the road."

The company is, as the source puts it, selling the picks and shovels while also positioning itself as the sheriff who keeps the gold rush from turning into a riot. Whether an open-source guardrail framework sees real adoption remains unproven; the announcement itself does the reputational work.

The takeaway: the security launch reads as narrative management timed to the buyback announcement.

Does this echo the dot-com era?

There is a historical parallel, and it is unflattering. By choosing to buy back its own stock rather than deploy that capital into more AI startups, Nvidia is effectively saying it has better uses for its cash.

"The more money that Nvidia sinks into startups, the less faith it has in their ongoing health," Zitron said. "Jensen is doing these investments to keep the ecosystem alive so that people keep buying GPUs."

William Lazonick, economist and professor emeritus at the University of Massachusetts, draws the direct comparison to the late-1990s. "In the internet boom of 1997-2000, this activity was called vendor financing as equipment providers funded service providers to purchase their equipment," he said. "Many of them subsequently went bankrupt. This time the funding is on a vastly increased scale."

The structural risk is familiar: when the vendor's revenue depends on customers who can only buy because the vendor or its allies fund them, demand can evaporate quickly. Nvidia's advantage is that it collects payment up front, whoever ultimately absorbs the losses. The buyback keeps the fear-of-missing-out narrative alive even as the underlying financials of its customers weaken, per the source.

The takeaway: the pattern mirrors vendor financing from 1997-2000, at far larger scale — and Lazonick notes many of those funded companies went bankrupt.

Who gains from the buyback?

Large tech companies routinely spend heavily on repurchases. "That's what shareholders expect — and they'll be penalized if they're not doing like their peers," Marx said.

The mechanics favor concentration. When Nvidia buys back shares, it reduces the number of shares on the market, pushing up the price of those that remain. In theory, every shareholder benefits. In practice, the gains skew heavily toward the largest holders and corporate insiders rather than people who own Nvidia indirectly through an index fund or retirement account. A 401(k) balance may tick up slightly; the scale of ownership determines who captures the real gains.

Lazonick goes further, arguing buybacks are the primary mechanism funneling profits from America's most successful companies to the richest 0.1%, giving them more capital "to engage in various forms of predation to make more money."

The source's own verdict is unsparing: Nvidia's historic authorization is another form of AI theater protecting those already at the top, while everyone else watches and wonders how long the music keeps playing. That characterization is an opinion, but the numbers behind it — $235 billion in authorized repurchases against an ecosystem still unable to answer a one-developer website's question about profitability — are the documented record.

The takeaway: buybacks mechanically reward concentrated ownership, and retail investors see the smallest share of the benefit.

What happens next?

The pressure point to watch is 2026, when roughly $300 billion of direct bond issuance from hyperscalers is expected to hit credit markets, according to the source. If Nvidia's customers must keep borrowing to keep buying, the buyback authorizations will tell investors whether Nvidia's board still believes the cash flows — or whether, as Zitron cautions, an authorization was only ever a headline.