Nvidia's $81.6B Quarter and $43B Startup Bet Lay Bare AI Capex

The News

Nvidia closed its fiscal Q4 with $81.6 billion in revenue, a 20% sequential lift, alongside a quieter disclosure that may prove the more telling one: its equity portfolio in privately held startups nearly doubled in three months, climbing from $22 billion in January to $43 billion in April. The chipmaker pumped $18.5 billion into private rounds during the quarter alone, according to its filing.

Data centre sales hit a record $75.2 billion, dominating the print. CFO Colette Kress told analysts that Blackwell had been "adopted by every major hyperscaler, cloud provider, and model maker." CEO Jensen Huang flagged Anthropic as a notable new buyer, telling investors the capacity coming online for the model lab "this year and next year will be quite significant" and that prior coverage of Anthropic "had been largely zero until this."

Forward guidance, however, marked the first deceleration in two years. The company expects $91 billion next quarter, implying just 12% growth and a sharp step down from the pace investors have become used to. Kress also flagged uncertainty around the H200 China shipments, where U.S. export approval has yet to translate into actual revenue.

Why It Matters

The headline figure is striking. The more consequential disclosure is the $43 billion private portfolio, an amount that now rivals the corporate venture books of Microsoft and Google. That figure excludes Nvidia's public-market positions in Corning and IREN, and ignores the separate $30 billion OpenAI commitment announced in February. The actual exposure to the build-out, when those are layered in, is substantially larger.

The last time a single supplier wielded this kind of platform-and-purse leverage was Cisco during the late-1990s telecoms boom, when its router franchise was reinforced by aggressive equity stakes in customers. Cisco's portfolio peaked just before the bust. The parallel is not exact, but the structural shape, chipmaker funds startup, startup buys chips, chipmaker books revenue and capital gain, is hard to miss.

The slowing forward guide is the bigger signal for capex watchers. Hyperscalers and model labs have been the entire growth engine, and a 12% quarter-on-quarter print, strong in absolute terms, suggests the easy doublings are done.

Indian Angle

For India, the immediate read is in capacity allocation. Nvidia is the principal supplier behind the IndiaAI Mission's compute pool, Reliance's planned AI data centre in Jamnagar, and the commercial GPU clouds at Yotta and CtrlS. Any global tightening, and the H200 China overhang almost guarantees re-routing, flows through India's queue first. Players that have already secured 2026 delivery slots, primarily Reliance, Tata and the IndiaAI consortium, just got a more valuable asset on their books.

The $43 billion portfolio also matters at the startup layer. Indian model labs Sarvam AI and Krutrim, both raising at multi-hundred-million-dollar valuations, are likely targets for the next round of strategic cheques. Nvidia has historically used such stakes to seed sovereign-AI plays in markets where it wants long-tail demand, and India, with its government-funded compute layer and English-speaking developer base, fits that template.

For Indian listed equities, the deceleration matters more than the headline beat. Names that have run on AI-supply narratives, including Persistent Systems, Tata Elxsi and Dixon, will be re-rated against the new 12% guide. The story is no longer Nvidia accelerating forever.

FAQ

When does this take effect?

The earnings cover the quarter ended late April 2026. The $91 billion forward guidance applies to the current quarter, which closes in late July, with results due in August. Portfolio holdings are point-in-time as of the filing date.

How does this compare to last quarter?

Nvidia booked roughly $68 billion in revenue the previous quarter. The 20% sequential jump to $81.6 billion is its strongest absolute dollar growth on record, yet the projected step up to $91 billion is the slowest sequential pace in two years.

What does this mean for Indian AI startups?

Sarvam, Krutrim and CoRover sit in the precise bucket Nvidia has been writing strategic cheques into. Expect approach calls. It also means GPU pricing for Indian buyers stays elevated through 2027, because no meaningful spare capacity is coming online.

Where can I read the original announcement?

The detailed filings are on Nvidia's investor relations site. TechCrunch's full coverage is linked in the source paragraph below.

This story was reported by TechCrunch. Read the full original coverage at TechCrunch.