Chip design
Designing the processors that train and run AI: Nvidia's GPUs, and custom chips built for Google, Amazon and others.
Who is making money
The companies earning or raising real money in this part of AI. Suppliers are marked: they sell to the builders rather than to end users.
the default AI chip
$89.0B data centre revenue in the quarter to July 2026 (fiscal Q2 2027), out of $96.2B total
Open profileIs designing its own AI chip with Broadcom, to run beside its Nvidia and AMD systems.
About $13B revenue in 2025, against about $34B of costs (reported)
Open profileits own AI chips
25% of AI chip supply (Epoch AI)
Open profileits own AI chips
5% of AI chip supply (Epoch AI)
Open profiledesigns custom chips for the big clouds
AI chip revenue $16.7B in a quarter, up 221%
Open profileMakes fast inference chips for running models, built around its own LPU design.
$3.5B in a 2026 round that followed its Nvidia licensing deal
Open profileMakes very large single-chip AI systems built for fast answers from models.
$510M in 2025, with $87.9M net income
Open profileDesigns custom AI chips for large cloud customers, and the optical chips that link data centres.
$2.006B in the quarter to July 2026 (fiscal Q2 2026), up 58%
Open profileLicenses the processor designs (CPU cores) that most of the world's chips are built on, and earns royalties on each chip sold.
$4.92B for the year to March 2026, a record
Open profileSells chip-design software and design blocks used to build AI and other chips.
$1.58B in the quarter to June 2026, up 24.2%. Full-year 2026 guidance about $6.3B
Open profileSells the software that engineers use to design and check AI chips, plus the chip building blocks they reuse.
$2.477B in the quarter to July 2026 (fiscal Q3 2026), up from $1.740B a year earlier. Full-year guidance about $9.7B
Open profilethe main GPU rival
Deal to supply up to 6 GW of chips to OpenAI from late 2026
Open profileDesigns AI chips and systems for running large models for companies and governments.
About $11B after a round of about $1B (July 2026)
Open profileDesigns chips built only for running transformer models, the design behind most current AI.
$500M Series B round in January 2026
Open profileDesigns AI chips and open source chip designs, led by chip designer Jim Keller.
Open profileWhere the money flows
Clouds, AI labs and GPU clouds pay Nvidia, AMD and Broadcom, who pay TSMC and the memory makers to build the chips.
pays Broadcom to design and build its custom AI accelerators, 10 GW from the second half of 2026 to 2029
financemagnates.cominvests in OpenAI as Nvidia systems go live, under a letter of intent for at least 10 GW
datacenterdynamics.combuys Nvidia systems for AI data centres under the same 10 GW letter of intent
datacenterdynamics.compays for a non-exclusive licence to Groq's chip technology and hires most of its executives
dealroom.cobuys TPU capacity from Google, up to one million chips with over 1 GW in 2026
datacenterdynamics.combuys next-generation TPU capacity built with Broadcom, starting in 2027
futurumgroup.com
The openings
Gaps still open in this area, from ones a small business could take to ones only giants can play.
- Small business Tracking cost per token and benchmarks across chip types for buyers
- Startup money Software that makes non-Nvidia chips easy to use (compilers, porting tools)
- Startup money Chips built only for running models (inference)
- Giants only Taking on Nvidia head to head
A handful of buyers account for most chip revenue, and some deals are circular: analysts doubt OpenAI can pay for its AMD order.
Sources: epoch.ai, runtimewire.com, fourweekmba.com, constellationr.com
The map updates three times a week.
241 companies across 25 areas, 62 money flows and 89 openings so far. New research lands every Tuesday, Thursday and Saturday.
Back to the full mapFree daily email
Get the map updates in your inbox
A short read every morning on what is moving in AI, including new areas of the map as they go live. Free.
You are in
Your first AI Pulse lands tomorrow morning. Keep an eye on your inbox.