Nvidia's Growth Story Is Getting Complicated — and the Market Is Paying Attention

Nvidia's week: DOJ reviews Groq licensing deal, talks over a $10B Anthropic IPO stake, and plans for 2GW of AI capacity in Australia by 2027.

Nvidia spent the past week proving that scale cuts both ways. On one hand, the chipmaker unveiled a sweeping build-out of AI infrastructure, deepened a strategic alliance with Palantir, and pushed further into consumer hardware. On the other, it confirmed that U.S. antitrust authorities are examining one of its licensing arrangements — and Reuters reported that a multi-billion-dollar investment in one of its own customers is still being weighed.

Washington Zeroes In on the Groq Arrangement

The company acknowledged that federal regulators are reviewing its licensing agreement with AI startup Groq. According to Reuters, the Justice Department is looking into whether the deal's structure was designed specifically to sidestep antitrust review. Nvidia has not disclosed details about the scope of the inquiry, and the matter is likely to shadow the company for some time.

That regulatory cloud arrives alongside another development that raises questions about how Nvidia funds its own growth. Reuters reported that the chipmaker is negotiating to serve as an anchor investor in Anthropic's planned initial public offering, with a potential commitment of up to $10 billion. Talks remain in the discussion phase, and no final decision has been made.

Should the investment proceed, it would tighten the already-intertwined relationship between Nvidia and one of the largest AI model developers — a pattern visible across several of the company's partnerships. For shareholders, the combination of regulatory pressure and escalating capital commitments to customers poses a fundamental question: how durable is Nvidia's business model when growth is increasingly purchased through stakes in its own buyers?

Australia Becomes the Blueprint for a Physical Bottleneck

Away from the legal and financial maneuvering, Nvidia is pressing ahead with bricks-and-mortar expansion. Together with Australian cloud and data-center partners, the company announced plans to build as much as 2 gigawatts of AI computing capacity in the country by 2027. The alliance includes Firmus, Sharon AI, IREN, ResetData, Megaport, CDC, NEXTDC and AirTrunk — a broad coalition expected to supply land, power and building capacity for Nvidia's DSX AI factories.

The message behind the Australian push extends well beyond the continent. Speaking at Goldman Sachs' Communacopia + Technology conference, CEO Jensen Huang made the case that the AI build-out is still in its early innings, with demand from cloud providers, enterprises, device makers and so-called neoclouds showing no signs of letting up. At the same time, he named the bottlenecks plainly: land, energy and supply chains. That kind of candor is notable from a chief executive whose stock trades on boundless growth expectations — and it suggests the company is taking its physical constraints more seriously than it did a year ago. Building data centers in Australia is not a prestige exercise; it reflects tightening capacity in established markets.

Palantir, Consumer PCs and the Broadening Portfolio

Nvidia also disclosed a collaboration with Palantir Technologies aimed at bringing "sovereign AI" into critical supply chains, beginning with Nvidia's own operations. The announcement reads as technical on its face, but carries strategic weight: Nvidia is positioning itself not merely as a chip supplier but as a partner for governments and security-sensitive infrastructure. For anyone tracking the geopolitical debate over AI sovereignty, it is a building block that draws the company closer to states and critical industries — and reduces its reliance on any single large customer, even as that customer concentration has repeatedly been flagged as a risk.

The company's consumer and creative ambitions are advancing in parallel. Lenovo and Acer are set to launch the first Windows PCs powered by the RTX-Spark chip in October, Nvidia told Reuters. On the software side, GeForce NOW is adding 26 titles in September, led by NBA 2K27 featuring the new DLSS-5 technology. Meanwhile, Nvidia is showcasing real-time AI for broadcast and streaming applications at IBC in Amsterdam and presenting local AI initiatives at IFA. None of these moves cannibalizes the others — a breadth that should make the company more resilient if the data-center boom ever cools.

What the Tape Says

The market has greeted this flood of news with restraint. Nvidia closed Friday at EUR 188.22, essentially flat versus the prior session. Over the past seven trading days the stock has shed 5.1%, and on a monthly basis it is down 3.3%. The shares sit roughly 7.1% below their 52-week high of EUR 202.50, reached in May.

The longer view remains firmly positive: the stock is up 17% since the start of the year and has gained 25% over twelve months. A Relative Strength Index reading of 48.9 puts the shares in neither overbought nor oversold territory — a sign that investors are balancing conflicting signals from the regulatory review, the billion-dollar investment plans and the operational expansion rather than taking a decisive stance.

The Takeaway

On balance, the positives carry more weight. Physical capacity building, strategic diversification through Palantir, and a broad presence across consumer and creative markets paint a picture of a company actively managing its dependencies rather than ignoring them. Huang's willingness to speak openly about land and energy constraints suggests management has a realistic read on risk — preferable to blind growth rhetoric. The recent softness in the share price looks more like general nervousness around AI valuations than a verdict on these specific initiatives.