Is Nvidia profitable in 2026?

Yes. Nvidia posted FY2026 (ended January 25, 2026) revenue of $215.9 billion, up 65% year over year, with net income around $120 billion. Q1 FY2027 (reported April 28, 2026) hit a record $81.6 billion revenue, up 85% YoY, with Data Center revenue alone at $75.2 billion, up 92%.

Could custom AI chips replace Nvidia GPUs?

Partially, in inference. Hyperscalers like Google, Amazon, Microsoft, and OpenAI (with Broadcom's 'Jalapeño', unveiled June 24, 2026) are shipping in-house silicon, and analysts project Nvidia's inference share could fall from roughly 90% to 20-30% by 2028. Training workloads still rely heavily on Nvidia as of mid-2026.

Who is Nvidia's biggest competitor?

Its own largest customers. Google (TPU v7), Amazon (Trainium3), Microsoft (Maia 200), Meta (MTIA), and OpenAI (Jalapeño, June 2026) are building custom AI chips while also buying Nvidia GPUs. Three direct customers made up 54% of Nvidia's total sales as of Q1 FY2027 (April 28, 2026).

Is Nvidia a safe company to join in 2026?

By the numbers, yes: Nvidia reported about 42,000 full-time employees at FY2026 close (January 25, 2026), up 16.67% year over year, with no layoffs disclosed. Regulatory risk exists, including a China SAMR antimonopoly preliminary finding since September 2025 and a French CUDA lock-in probe reported in July 2026.

Nvidia is publicly traded (NASDAQ: NVDA) with no controlling parent or private owner. It became the world's most valuable public company, with market cap reaching roughly $4.9 to $5.1 trillion in July 2026 and briefly touching $5.42 trillion intraday on May 19, 2026.

Is Nvidia exposed to China risk?

Significantly. The H20 chip was export-restricted in April 2025, forcing a $4.5-5.5 billion inventory charge. H200 chips were conditionally cleared for about 10 Chinese firms by May 2026, with actual shipments only beginning mid-July 2026 amid customs friction, and Nvidia's China/Hong Kong revenue was just 9.1% of FY2026 sales.

AI companies and labs · United States

Nvidia

nvidia.com ↗

Nvidia makes the GPUs and software stack that train and run most AI models today.

Defensible for now

Assessed 20 July 2026 · Evidence as of 20 July 2026

Real protection, with named pressure against it.

On the recordSince 20 July 2026 · Checkable from August 26, 2026 · Every call on the record Checkpoint: Q2 FY2027 earnings report, first real read on custom silicon's dent in Data Center growth

The call

Nvidia is still growing faster than almost any large company in history, even as the companies writing its biggest checks quietly build ways to need it less.

Nvidia makes the specialized computer chips that power artificial intelligence systems, and it currently sells most of them to companies like Microsoft, Google, and Amazon. It is hard to replace right now because its chips work really well together through special software, and Nvidia has long-term deals with the factory that makes them. However, the same companies buying from Nvidia are spending billions to design their own chips for their AI work, which could cut Nvidia's sales roughly in half by 2028, especially for the less demanding tasks. The thing that matters most: whether these customer-built chips actually start replacing Nvidia chips in real datacenters by the middle of 2026, which would show up as slower growth in Nvidia's sales to these customers.

Why

AI pressure: Custom AI chips from Google TPU, Amazon Trainium, Microsoft Maia, Meta MTIA, OpenAI Jalapeño. Nvidia's own top customers are building their own chips to run AI cheaper, aiming at the inference workloads Nvidia dominates today. Could AI do the core job? Only partly.

Bigger players: Microsoft, Google, Amazon, Meta, OpenAI. These five buy over half of Nvidia's chips while also funding rival silicon, making Nvidia's biggest customers its biggest long-term risk. Could a giant copy it? Probably.

What protects it. CUDA software and NVLink networking keep AI labs locked into Nvidia's full hardware and code stack, hard to swap out mid-project.

What would change this call

Nvidia Data Center revenue growth rate in Q2 or Q3 FY27 earnings reports. A drop below 60% YoY growth would show custom silicon is eating into demand.
Public disclosure of AWS Trainium3, Google TPU v7, or OpenAI Jalapeño training (not just inference) deployments. Would show custom chips moving beyond inference into training, Nvidia's core stronghold.
Outcome of French antitrust complaint over CUDA lock-in and CoreWeave equity stakes. Would show regulatory pressure on Nvidia's software ecosystem moat.

Key numbers

FY2026 revenue $215.9B, up 65% year over year (2026-01-25)
Q1 FY2027 Data Center revenue $75.2B, up 92% (2026-04-28)
Three customers make up 54% of total revenue (2026-04-28)

Named risks

54% of revenue concentrated in three customers who are simultaneously building competing chips. Severity: high
Analysts project inference market share erosion from ~90% toward 20-30% by 2028. Severity: high
China export control volatility and unresolved SAMR antimonopoly finding since September 2025. Severity: medium
French antitrust complaint over CUDA lock-in reportedly being prepared as of July 2026. Severity: medium
Circular hyperscaler capex and $662B off-balance-sheet lease exposure flagged by Moody's, May 2026. Severity: medium

Rivals, as read in July 2026

Rival	Direction	Evidence
Google (TPU)	gaining on	TPU v7 Ironwood/Trillium shipping to internal and cloud workloads as of 2026
Amazon (Trainium/Inferentia)	gaining on	Trainium3 and Inferentia ramping for inference workloads through 2026
Microsoft (Maia)	contesting	Maia 200 in deployment while Microsoft remains a top Nvidia customer, 2026
OpenAI/Broadcom (Jalapeño)	gaining on	In-house AI chip unveiled June 24, 2026, direct customer-to-competitor shift
AMD	contesting	Instinct MI350X/MI450 positioned as merchant-silicon alternative through 2026
Huawei	gaining on	Ascend 910C/950PR dominant in China post export ban, 2025-2026

Questions people ask

Is Nvidia profitable in 2026?: Yes. Nvidia posted FY2026 (ended January 25, 2026) revenue of $215.9 billion, up 65% year over year, with net income around $120 billion. Q1 FY2027 (reported April 28, 2026) hit a record $81.6 billion revenue, up 85% YoY, with Data Center revenue alone at $75.2 billion, up 92%.
Could custom AI chips replace Nvidia GPUs?: Partially, in inference. Hyperscalers like Google, Amazon, Microsoft, and OpenAI (with Broadcom's 'Jalapeño', unveiled June 24, 2026) are shipping in-house silicon, and analysts project Nvidia's inference share could fall from roughly 90% to 20-30% by 2028. Training workloads still rely heavily on Nvidia as of mid-2026.
Who is Nvidia's biggest competitor?: Its own largest customers. Google (TPU v7), Amazon (Trainium3), Microsoft (Maia 200), Meta (MTIA), and OpenAI (Jalapeño, June 2026) are building custom AI chips while also buying Nvidia GPUs. Three direct customers made up 54% of Nvidia's total sales as of Q1 FY2027 (April 28, 2026).
Is Nvidia a safe company to join in 2026?: By the numbers, yes: Nvidia reported about 42,000 full-time employees at FY2026 close (January 25, 2026), up 16.67% year over year, with no layoffs disclosed. Regulatory risk exists, including a China SAMR antimonopoly preliminary finding since September 2025 and a French CUDA lock-in probe reported in July 2026.
Who owns Nvidia?: Nvidia is publicly traded (NASDAQ: NVDA) with no controlling parent or private owner. It became the world's most valuable public company, with market cap reaching roughly $4.9 to $5.1 trillion in July 2026 and briefly touching $5.42 trillion intraday on May 19, 2026.
Is Nvidia exposed to China risk?: Significantly. The H20 chip was export-restricted in April 2025, forcing a $4.5-5.5 billion inventory charge. H200 chips were conditionally cleared for about 10 Chinese firms by May 2026, with actual shipments only beginning mid-July 2026 amid customs friction, and Nvidia's China/Hong Kong revenue was just 9.1% of FY2026 sales.

Timeline

2026-07-19Market cap hits ~$5.1T; Japan deal for 27,500 Rubin GPUs; Dell $5B Rubin deal announced
2026-07-14First H200 GPU shipments to China begin after conditional clearance for ~10 firms
2026-07-12French antitrust authority reportedly preparing complaint over CUDA lock-in, CoreWeave stakes
2026-06-24OpenAI unveils in-house 'Jalapeño' AI chip built with Broadcom
2026-06-01CoreWeave validates first Vera Rubin NVL72 rack
2026-05-20Q1 FY2027 results: $81.6B revenue (+85% YoY); dividend raised from $0.01 to $0.25/share
2026-03-16Vera Rubin expands to seven-chip platform at GTC, reaches full production by June 2026
2026-02-25FY2026 results: $215.9B revenue (+65% YoY), ~$120B net income
2026-01-05Vera Rubin AI platform unveiled at CES with six chips
2025-12-31Nvidia acquires Groq's LPU inference technology for roughly $20 billion

Revision history

20 July 2026

First assessed. Defensible for now. The text above is fixed as of this date and is not edited afterwards.

In the same band

Broadcom Marvell Technology Advanced Micro Devices