Nvidia’s Two-Front War: Pentagon AI Access and a $1 Trillion Order Backlog

Nvidia secures Pentagon AI contract and faces $1 trillion in orders as hyperscaler capex surges 77%, shifting its narrative from hardware to strategic systems partner.

The narrative around Nvidia is shifting from pure hardware dominance to something far more strategic. While the chipmaker’s stock has been taking a breather—closing Friday at roughly $198, marking four consecutive losing sessions—the underlying business is being reshaped by two powerful forces: a landmark Pentagon deal and an unprecedented wave of hyperscaler spending that has pushed order books past the trillion-dollar mark.

The most transformative development came on May 1, when Nvidia and Alphabet, alongside other tech heavyweights, signed agreements with the U.S. Department of Defense to deploy their AI models on classified military networks. This isn’t about selling chips to the Pentagon. Nvidia’s “Nemotron” AI models—specialized software agents capable of operating autonomously on secure systems—are the real prize. The move elevates Nvidia from a component supplier to an integrated systems partner for the U.S. armed forces, opening a long-term, high-retention revenue stream that analysts say could fundamentally alter the company’s risk profile.

Yet the Pentagon’s embrace comes with a geopolitical shadow. Senator Chris Coons has pressed Commerce Secretary Lutnick for explanations regarding comments about H200 chip exports to China, keeping trade tensions alive. These restrictions remain a persistent headwind for Nvidia’s revenue mix, though analysts have largely brushed them aside. The consensus rating remains a “Strong Buy,” with an average price target of $266—roughly 34% above current levels.

The Hyperscaler Spending Tsunami

While the defense angle is new, the core driver of Nvidia’s business remains the insatiable appetite of the world’s largest tech companies. Google, Amazon, Microsoft, and Meta are planning combined capital expenditures of roughly $725 billion for 2026—a 77% surge from the prior year and $100 billion more than was projected just last quarter. Microsoft alone is earmarking $190 billion in investments this year and has warned of capacity constraints by the end of 2026, signaling structurally tight GPU availability.

Nvidia CEO Jensen Huang has put a staggering number on the demand: purchase orders for the company’s next-generation chip families, including the Vera-Rubin platform due for first customer shipments in the second half of 2026, total $1 trillion by the end of 2027. The Vera-Rubin architecture promises ten times the performance per watt over its predecessor, a critical metric as energy efficiency becomes the primary constraint in data center expansion.

The company’s fiscal 2026 results underscore the momentum: revenue surged roughly 65% to around $216 billion, with profits rising in similar proportion. For the current quarter, management has guided for 77% revenue growth, while analysts are penciling in 79%. The options market is pricing in a roughly 10% swing in the stock by the end of May, with the data center segment’s revenue being the make-or-break metric.

The Competitive Landscape Tightens

Nvidia’s dominance is not going unchallenged. AMD is gaining ground with its own AI chips, and major customers like Amazon are developing custom accelerators for specific workloads. AMD’s stock hit a new 52-week high on Friday at €302.15, up more than 58% year-to-date and roughly 246% over the past twelve months. The company reports Q1 earnings on Tuesday, with analysts expecting revenue of nearly $9.85 billion—a 32% year-over-year increase—and earnings per share of around $1.27.

The margin story at AMD is more nuanced. The company expects a gross margin of roughly 55% in Q1, two percentage points lower than the prior quarter, due to the removal of a one-time inventory liquidation benefit tied to MI308 sales to China. Those China-related revenues have collapsed from an estimated $390 million in Q4 to just $100 million. However, the pipeline remains robust: AMD has secured a 6-gigawatt deal with OpenAI, with first Instinct MI450 GPUs expected in the second half of 2026, and a similar 6-gigawatt partnership with Meta featuring custom MI450-based GPUs and sixth-generation EPYC CPUs.

Alphabet, meanwhile, is making its own play. The Google parent posted a 22% revenue increase to $109.9 billion in its latest quarter, with Google Cloud accelerating to 63% growth and hitting $20 billion in revenue. CEO Sundar Pichai noted that enterprise AI solutions have become the primary growth driver in the cloud business. Alphabet has raised its 2026 investment plan to between $180 billion and $190 billion and is developing new custom TPU chips—the TPU 8t and TPU 8i—with plans to sell them to select third-party customers for the first time. That is a direct assault on Nvidia’s GPU stronghold.

The Infrastructure Bottleneck

The physical constraints of building out AI infrastructure are becoming a story in themselves. ASML, the Dutch lithography monopoly, reported Q1 net sales of €8.77 billion and net profit of €2.76 billion, with a gross margin of 53%. CEO Christophe Fouquet stated that “demand for chips exceeds supply,” and the company has raised its full-year guidance to between €36 billion and €40 billion in revenue. ASML plans to deliver 60 of its best-selling low-NA EUV systems in 2026, a 25% increase, and aims to scale capacity to 80 machines by 2027.

China’s share of ASML’s sales has shrunk to 19% in Q1 from 36% in the prior quarter, with South Korea becoming the largest market at 45%. The proposed MATCH Act could impose further export restrictions, potentially hitting the lucrative service business, which accounts for roughly 28% of total revenue from an installed base of over 5,000 machines.

SoftBank is tackling the infrastructure bottleneck from a different angle. The Japanese conglomerate is working on a new entity called “Roze,” which would use autonomous robots to build data centers. The U.S. is short an estimated 439,000 construction workers, and satellite imagery suggests up to 40% of AI data center construction sites are experiencing delays. Roze would bundle existing energy, land, and infrastructure assets from SoftBank’s portfolio with ABB Robotics, whose acquisition SoftBank agreed to last year. The company could go public in the second half of 2026 with a valuation of around $100 billion—though both the valuation target and timeline are considered ambitious internally, not least due to geopolitical uncertainties in the Middle East.

The $5 Trillion Question

Alphabet’s stock surged 9.58% on Friday to a new all-time high of €328.30, pushing its market capitalization toward $5 trillion and putting it in striking distance of Nvidia’s crown as the world’s most valuable company. The cloud business’s disclosed backlog of $467.6 billion, overwhelmingly tied to Google Cloud, underscores the depth of future demand.

For Nvidia, the immediate test comes on May 20, when the company reports quarterly results. The biggest risk flagged by analysts is a potential dip in hyperscaler investment ahead of the new chip rollout—a classic pause-before-the-next-cycle pattern. But with $1 trillion in orders on the books and a new Pentagon partnership that locks in government revenue, the company’s long-term trajectory appears firmly intact.

The next phase of the AI cycle will be defined not just by technological leaps, but by geopolitical maneuvering, infrastructure bottlenecks, and the race to build the physical foundation of the revolution. Nvidia is betting it can win on all fronts—and the market is watching to see if it can deliver.