Nvidia's Hyperscaler Dilemma: Big Customers Turn Rivals as AI Chip Dominance Faces Its Sternest Test

Nvidia faces a turbulent week as Google and Amazon pivot to selling their own AI chips externally, threatening its 90% market share amid $725B hyperscaler spending.

Nvidia is navigating one of its most turbulent weeks in recent memory, with the stock swinging between competing narratives of unshakeable demand and mounting competitive threats. The chipmaker's shares closed Thursday at €171.10, down roughly 4.5%, as investors digested a strategic pivot by its biggest customers that could reshape the semiconductor landscape.

For years, Google and Amazon were Nvidia's most loyal buyers, snapping up every available AI accelerator to power their cloud empires. Now those same hyperscalers are signaling a dramatic shift. Both companies have reportedly decided to sell their in-house processors—Google's TPUs and Amazon's Trainium chips—directly to external customers, ending a long-standing policy of keeping that hardware exclusive to their cloud platforms.

The implications for Nvidia are profound. The company currently commands an estimated 90% of the global market for AI accelerators, the essential building blocks for the data center expansion powering the artificial intelligence boom. Market observers view the hyperscalers' opening of their proprietary hardware as an irreversible trend, driven by a collective desire to reduce the industry's dependence on a single external supplier.

The $725 Billion Question

The scale of the challenge is matched only by the size of the budgets involved. Meta, Alphabet, Microsoft, and Amazon have committed to spending a combined $725 billion on infrastructure by the end of 2026. Alphabet alone is targeting capital expenditures of up to $190 billion, while Meta has raised its investment ceiling significantly.

Yet a growing chorus of skepticism is emerging about the profitability of these colossal outlays. Investors are repositioning their portfolios, weighing how long Nvidia can sustain its extraordinary margins when cost-conscious enterprise customers begin migrating to the alternative architectures being developed by the very companies that were once its biggest clients.

The stock's reaction reflects this tension. Despite Thursday's pullback, Nvidia shares remain up more than 78% year-to-date, trading just shy of the 52-week high of €182.26 reached only days ago. The long-term uptrend remains intact, but the path forward looks increasingly complicated.

A Legal AI Bet and a Price Surge in China

Amid the strategic headwinds, Nvidia continues to execute on multiple fronts. On April 30, the company confirmed alongside Atlassian a $50 million extension of the Series D funding round for Legora, a legal AI startup that automates document processing and legal analysis. The investment values the platform at $5.6 billion and fits a familiar pattern: Nvidia is placing bets on software companies that build on its hardware, locking in future demand for its accelerators.

The strategy is already paying dividends in China, where scarcity is driving prices to extraordinary levels. A single B300 chip based on the Blackwell architecture now commands nearly 7 million yuan—roughly double its previous price. Chinese technology companies continue to scramble for high-performance GPUs despite US export restrictions, while authorities crack down on informal smuggling networks. The result is a market where Nvidia's pricing power remains formidable, even under geopolitical constraints.

Fresh Leadership and a Key Earnings Date

On the personnel front, Nvidia is bringing in outside experience. Scott Gawel, formerly of Intel, will assume the role of Chief Accounting Officer on May 4, 2026. His compensation package includes an $800,000 annual salary plus restricted stock units valued at approximately $12.9 million, vesting over four years.

All eyes are now on May 20, when Nvidia's management is scheduled to report first-quarter fiscal results. The market will be scrutinizing initial sales figures for the new Blackwell architecture, looking for evidence that demand for Nvidia's high-performance chips remains robust enough to fend off the encroachment from its own biggest customers.

The stock briefly dipped mid-week on reports that OpenAI missed certain user growth and revenue targets in late 2025, reigniting doubts about whether the industry's massive AI infrastructure spending will generate returns quickly enough. But the shares have since stabilized, trading at €179.42—roughly 19% above their level 30 days ago and well above all key moving averages.

Nvidia's long-term ambition remains undimmed. The company has set a target of generating $1 trillion in cumulative revenue from the Blackwell and upcoming Vera-Rubin platforms by 2027. Whether that goal survives the hyperscaler rebellion will depend on the numbers delivered on May 20.