Nvidia Posts Record $81.6B Quarter as Jensen Huang Concedes China AI Market to Huawei

Nvidia delivered another record-shattering quarter on May 20, reporting $81.6 billion in revenue for Q1 of its fiscal year 2027 — an 85% year-over-year surge and a 20% sequential climb that once again eclipsed Wall Street’s expectations. But the headline number told only half the story. The following day, in a CNBC interview that reverberated across financial markets, CEO Jensen Huang said the quiet part out loud: Nvidia has “largely conceded” China’s advanced AI chip market to Huawei.

“Huawei is very, very strong,” Huang said. “Their local ecosystem of chip companies are doing quite well, because we’ve evacuated that market.”

The candid acknowledgment, delivered matter-of-factly rather than with defensiveness, marked a watershed moment — the definitive end of years of diplomatic hedging around Nvidia’s China exposure and a frank admission that US export controls have permanently reshaped the competitive landscape in the world’s second-largest AI market.

The Numbers That Stunned Wall Street

The Q1 FY2027 results — covering the period ended April 26, 2026 — were nearly flawless on paper. Revenue of $81.62 billion surpassed the consensus of $79.18 billion. Earnings per share came in at $1.87 versus the $1.77 analysts had forecast. Data center revenue reached $75.2 billion, up 92% year-over-year, cementing Nvidia’s position as the backbone of global AI infrastructure. Gross margins held at a formidable 74.9% on a GAAP basis.

The board sweetened the results with an $80 billion additional share repurchase authorization — one of the largest buyback expansions in corporate history — and raised the quarterly cash dividend from $0.01 to $0.25 per share, a 25-fold increase that reflects confidence in the company’s cash generation.

Forward guidance was similarly bullish. For Q2 FY2027, Nvidia projected revenue between $89.1 billion and $92.8 billion, versus the $87.3 billion Wall Street had expected — implying a sequential increase of up to 14%.

Yet Nvidia shares fell roughly 2% when markets opened and closed down 1.4% on the day. The paradox of the “sell the news” reaction on a spectacular beat has become a routine feature of Nvidia’s earnings cycles. The stock is priced for consistent miracles, and even miracles can disappoint at a high enough multiple.

The China Vacuum: A $15 Billion Annual Headwind

The more consequential story emerged not from the earnings release, but from Huang’s CNBC interview on May 21. His acknowledgment that Nvidia has “largely conceded” China’s AI chip market encapsulates years of escalating US export controls that have effectively locked the company out of a market that once generated roughly 20% of its data center revenue.

At the current data center run rate of $75 billion per quarter, China’s 20% historical share represents a potential $60 billion annual revenue pool that Nvidia cannot access. Even if Chinese demand has grown since those baseline measurements, the scale of the lost opportunity is staggering. The Trump administration’s April 2026 requirement that Nvidia obtain export licenses to sell chips to China — adding a new layer of bureaucratic friction on top of existing Blackwell restrictions — was the effective killing blow to any recovery scenario.

What makes Huang’s statement particularly significant is what it says about US policy outcomes. The export controls were designed to prevent Nvidia’s advanced AI accelerators from accelerating China’s military applications. They have succeeded in that narrow objective — but the unintended consequence has been to hand Huawei a near-monopoly on China’s rapidly expanding AI infrastructure market.

Baidu, Alibaba, Tencent, and ByteDance — the companies that would otherwise have been Nvidia’s largest Chinese customers — have all committed publicly to building their AI stacks on Huawei’s Ascend 910C and 910D processors. Local chip startups including Cambricon and Biren are filling out the surrounding ecosystem. The market that US policymakers hoped to contain has instead become a hothouse for domestic Chinese semiconductor development.

Huawei’s Unlikely Ascent

The speed of Huawei’s rise in AI chips is a direct function of the vacuum Nvidia created. Three years ago, the company’s AI chip business was considered a strategic liability — underpowered relative to Nvidia, dependent on restricted manufacturing at SMIC, and unable to match the CUDA software ecosystem that has given Nvidia a near-impenetrable developer moat in the West.

Today, that assessment is outdated. The Ascend 910C competes credibly with the H100 on training workloads for large language models. The 910D, announced in early 2026, targets H200-class performance. More importantly, the associated software stack — MindSpore for training, MindIE for inference — has matured dramatically as Chinese enterprises and researchers have had no choice but to adopt it.

Huang’s phrase “we’ve evacuated that market” is blunt but accurate. Nvidia did not lose China through competitive failure; it was forced out through policy decisions made in Washington. Whether those decisions were strategically wise — given that they created the very domestic Chinese AI chip champion they were meant to suppress — is a question policymakers have not yet publicly grappled with.

The Road Ahead: Vera Rubin and Sovereign AI

Nvidia’s answer to the China loss is to grow fast enough everywhere else that the gap becomes a rounding error. The Blackwell architecture currently powering deployments is ramping faster than expected, according to CFO Colette Kress. The next-generation Vera Rubin platform — named after the astronomer who discovered dark matter — remains on track for second-half 2026 delivery, with performance-per-watt roughly double that of Blackwell.

Vera Rubin is architecturally optimized for inference, not just training — a crucial distinction as AI deployment shifts from building models to running them at scale across millions of users. Inference workloads require different hardware characteristics: lower latency, higher throughput on smaller batch sizes, and the ability to run continuously at high utilization rather than in discrete training bursts.

Beyond hyperscalers, Nvidia is actively cultivating what Huang calls “sovereign AI” — the wave of national governments building their own AI infrastructure. Saudi Arabia, France, Japan, the UAE, and several other nations have committed to Nvidia-powered AI factories as strategic investments in their economic competitiveness. This long tail of sovereign customers diversifies Nvidia’s revenue base away from the four US hyperscalers that currently dominate its order book.

What Investors Are Actually Watching

With guidance of $89–$93 billion for the next quarter, the financial trajectory remains extraordinary. The concern isn’t the near-term numbers — it’s the competitive landscape in 18 to 24 months. AMD’s Instinct MI350X series has quietly gained traction in inference deployments. Google’s custom TPU v6 has allowed the company to reduce its Nvidia dependency. Intel, despite years of missteps, is still fighting to be relevant in AI accelerators.

None of these competitors pose an immediate threat to Nvidia’s dominance. But they represent a future in which the moat narrows — and the China concession is a reminder that even the deepest technical and ecosystem advantages can be overcome when the rules of engagement change.

For now, $81.6 billion says Nvidia is still winning. The question is which game it’s winning in, and whether the game is about to change.