Nvidia accelerators sit in 252 of the 500 systems on the November 2026 TOP500 list of the world’s most powerful supercomputers, released on 16 November at SC26 in Chicago, against 34 for AMD, by our count of the list’s spreadsheet. Two years earlier, in the November 2024 edition, the count was 183 for Nvidia and 19 for AMD.
The count is only half the picture. AMD’s 34 systems hold 28.9% of the list’s summed HPL performance and four of the top 10 places, against 44.4% and three places for Nvidia’s 252, by our arithmetic on the same spreadsheet. The average AMD system on the list delivers almost five times the HPL performance of the average Nvidia one, by our arithmetic.
Business Pill · GPUs explained
An evergreen Business Engineer explainer on GPUs. It teaches the general idea only and does not discuss the TOP500 list.
The key insight: As we read the spreadsheets, Nvidia has won the count on the TOP500 and AMD has won the biggest machines: 252 Nvidia-accelerated systems against 34, yet 28.9% of the list’s summed performance sits on AMD systems against 44.4% on Nvidia’s, by our arithmetic.
What the November List Shows
TOP500’s own release describes a quiet edition: “The ten leading systems did not change since the June edition”, and “Just 27 systems are new, the lowest number ever recorded by a wide margin”. LineShine, in Shenzhen, stays No. 1 at 2.198 exaflop/s, ahead of El Capitan at 1.809 exaflop/s, both as listed on the November list.
Accelerator-based systems rose from 275 to 291, or 58.2% of the list, the release says. The largest newcomer is HANGANG at the KISTI Supercomputing Center in South Korea, No. 12, which “combines NVIDIA Grace CPUs with GH200 Superchips and delivers 342.6 Petaflop/s on HPL”.

Count Versus Muscle
We counted a system as Nvidia or AMD when the spreadsheet’s accelerator field names that vendor. Nvidia’s count rose in each edition we checked: 183, 202, 219, 237 and 252, from November 2024 to November 2026. That is 15 more systems since June and 69 more since November 2024. AMD went from 19 to 34 over the same two years.
Measured by performance, the order is closer. Nvidia systems add up to 8.754 exaflop/s, or 44.4% of the list’s summed Rmax, and AMD systems to 5.688 exaflop/s, or 28.9%, by our arithmetic. That works out to about 34.7 petaflop/s for the average Nvidia system and 167.3 petaflop/s for the average AMD one.
The reason is at the top. El Capitan and Frontier, Nos. 2 and 3, use AMD Instinct accelerators, as do Eni’s HPC7 and HPC6 at Nos. 6 and 8. Nvidia’s top-10 places are JUPITER Booster at No. 5 and Alps at No. 10, both on GH200 Superchips, and Microsoft Azure’s Eagle at No. 7 on H100s.

What Nvidia Brought to SC26
NVIDIA used the conference to announce Rubin HPC, a GPU in the Vera Rubin platform for simulation and data processing, with a record number of CUDA cores, “more than 50,000 on a single GPU”; it says Rubin HPC “is expected to be available in 2028”. Ian Buck, vice president of hyperscale and high-performance computing at NVIDIA, said in the release: “For agents tackling science and engineering, faster simulation means faster discovery”.
Our earlier piece covered the rest of NVIDIA’s 16 November slate, including the Vera CPU and quantum releases. Nvidia reports third-quarter fiscal 2027 results after the US close on Tuesday 17 November, as our earnings preview set out, after the stock closed down 4.12% on Monday.
The Structural Read
As we read the five editions, Nvidia’s gains come from many mid-sized systems: its count rose by 69 in two years while its share of summed performance moved between 35.1% and 47.6%, by our arithmetic on the same spreadsheets.
As we read TOP500’s release, the list now says less about where accelerator demand is going: it counts only 27 new systems and itself points to AI factories as the place where growth has shifted.
TOP500.org, November 2026 release
“This suggests that growth has indeed shifted toward AI factories, and from capability systems to capacity systems.”
Three Implications
FOR NVIDIA Its accelerators are in 252 of the 500 systems, and HANGANG, the largest newcomer at No. 12, uses Grace CPUs with GH200 Superchips, the TOP500 release says.
FOR AMD Its Instinct accelerators are in El Capitan and Frontier at Nos. 2 and 3 and in Eni’s HPC7 and HPC6 at Nos. 6 and 8, the November list shows.
FOR THE TOP500 ITSELF The release says just 27 systems are new, the lowest number ever recorded, and the top 10 is unchanged since June.
The Business Engineer Lens
This story maps onto the Business Engineer framework The AI Supercycle.
The framework says: “Three competitive categories within one layer: generalist substrate (NVIDIA), hyperscaler custom (TPU, Trainium, MTIA), merchant custom challengers (Cerebras, Groq, SambaNova).”
As we read the TOP500 counts, the generalist position shows up as breadth, 252 systems across many sites, while AMD’s place on the list rests on a few of the largest machines; the framework predates the list and does not discuss it.
What Is Not Established
The system counts and shares are our arithmetic on TOP500’s spreadsheets, using the accelerator field only, so a system whose Nvidia parts are listed elsewhere, such as a machine that lists Nvidia only as its CPU, is not in our Nvidia count. The spreadsheet gives 292 entries with an accelerator and a summed Rmax of 19.70 exaflop/s, against the release’s 291 and 19.57; our shares use the spreadsheet’s own column.
The TOP500 measures HPL results that sites choose to submit. It is not a sales or market-share figure, and the release itself, citing Epoch AI figures, says “growth has indeed shifted toward AI factories”. NVIDIA’s releases we read give no TOP500 count of their own.
The Bottom Line
Nvidia accelerators are in 252 of the 500 systems on the November 2026 TOP500, up from 183 two years ago, against 34 for AMD, by our count. AMD’s fewer, larger machines still hold 28.9% of the list’s summed performance and four of the top 10, against 44.4% and three for Nvidia, by our arithmetic.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. The list results come from TOP500.org’s 16 November release and its November 2026 list page, which we read in full. The Nvidia and AMD system counts, performance shares and averages are our arithmetic on TOP500’s downloadable list spreadsheets for the November 2024, June 2025, November 2025, June 2026 and November 2026 editions, counting the accelerator field only. The Rubin HPC details and Ian Buck’s quote come from NVIDIA’s 16 November announcements.
The lead was a post on X about NVIDIA’s presence at SC26; we used NVIDIA’s and TOP500’s own publications instead. We did not contact Nvidia, AMD or TOP500 for this piece.
Sources: TOP500.org: November 2026 release (16 Nov 2026) · TOP500 list, November 2026 · TOP500 November 2026 list spreadsheet · NVIDIA: Rubin HPC announcement (16 Nov 2026) · NVIDIA: Vera CPU platform release (16 Nov 2026)









