The core of market concerns lies in Google's announcement to sell its self-developed TPU chips externally, while Amazon's self-developed Trainium chip has achieved an annualized revenue exceeding 20 billion US dollars. NVIDIA's largest customer is accelerating its transformation into a competitor. Analysts warn that this move could "fundamentally disrupt NVIDIA," though some argue that under the explosive growth of AI demand, supply remains the bottleneck.
The intensive increase in AI capital expenditures by tech giants should have been a significant boon for NVIDIA, but the market delivered an entirely opposite response.
On Thursday, $NVIDIA (NVDA.US)$ the stock price fell by more than 4%, breaking below the 200-dollar mark, with a single-day decline of nearly 10 dollars.
The previous evening, Meta, Alphabet, Microsoft, and Amazon successively released their earnings reports, with the four hyperscale cloud providers collectively projecting investments of up to $725 billion by 2026 for AI infrastructure construction.
NVIDIA holds approximately 90% of the AI accelerator chip market. Logically, this wave of investment should directly benefit the chip giant. However, investor concerns arise as NVIDIA’s most important clients begin large-scale development of in-house chips, potentially challenging its market dominance.
Previously, Alphabet announced that it would sell its self-developed TPU chips to external customers, while Amazon emphasized the rapid growth of its self-developed chip business during its earnings call.
Google TPU's "breakthrough" touches the market's most sensitive nerves.
Alphabet announced that it would sell its self-developed TPU chips to select external customers, who could deploy them within their own data center infrastructures.
Previously, TPUs were almost exclusively used within Google's internal ecosystem. Once commercialized externally, TPUs will evolve from being potential competitors to NVIDIA's GPUs into substantive market rivals.
Although TPUs are generally considered less versatile than NVIDIA’s solutions, they demonstrate notable cost-effectiveness advantages for specific AI application scenarios.
Amazon also highlighted the expansion momentum of its self-developed chip business during its earnings call.
According to Bloomberg, Amazon CEO Andy Jassy stated that the company's annualized revenue from its chip business has surpassed $20 billion, with triple-digit year-on-year growth. The core product is its self-developed Trainium chip.
Wall Street Analysts: The Wave of Self-Developed Chips Poses a 'Significant Risk'
Regarding this competitive landscape, analysts on Wall Street have issued clear warnings.
Seaport Research semiconductor analyst Jay Goldberg stated outright:
This could fundamentally disrupt NVIDIA; I consider it a fairly significant risk.
Goldberg’s reasoning is based on the scale and capital strength of hyperscale cloud providers. These companies are not only NVIDIA’s largest customers but also continue to invest resources in becoming its competitors.
Alphabet’s cloud revenue in the first fiscal quarter increased by 63% year-on-year to $20.03 billion, with backlogs nearly doubling to over $460 billion, a significant portion handled by internal TPUs.
Meanwhile, Meta raised its 2026 capital expenditure guidance to between $125 billion and $145 billion while simultaneously advancing its MTIA self-developed chip project.
However, not all analysts agree with this pessimistic assessment.
Stacy Rasgon of Bernstein Research countered, arguing that “focusing on who wins or loses is the wrong question.” His logic is that the rise of AI agents has caused computational demand to explode, with the current bottleneck for the industry being supply rather than demand.
Against this backdrop, all chip manufacturers with credible production capacity, including NVIDIA, are able to achieve full-capacity sales.
NVIDIA currently holds $95.2 billion in supply commitments, with cooperative clients including leading institutions such as OpenAI, Anthropic, CoreWeave, and Meta.
Editor/Jeffy