The Groq chip has 500MB of SRAM, while a Rubin chip has 288GB of memory.
Resolved true The record shows it happened or was so.
Factual: Statements about the present or the past, checked against the record: did they hold up?
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
Judge as of 2026-03-16. The claim has two parts: (a) the Groq chip (LPU) has about 500MB of on-chip SRAM; (b) an NVIDIA Rubin GPU has 288GB of memory (HBM4). True: official specifications support both figures, roughly (a Groq chip with about 500MB of SRAM, for example the current or next-generation LPU) and exactly as specified for Rubin (288GB). Mostly true: both are correct in substance with a small slip, for example Groq's SRAM is about 230MB on the first-generation chip but about 500MB on the relevant chip, or Rubin's memory is stated per package rather than per die. Mixed: one figure is correct and the other is substantially wrong. Mostly false or false: both are substantially wrong. Use false when neither is near the specification.
Set before the research began.
Evidence (4)
-
E1 How Nvidia's $20 billion Groq 3 LPU deal reshapes the Nvidia Vera Rubin Platform — Samsung 4nm process serves as bedrock for SRAM-based AI accelerator chip Tom's Hardware · 19 March 2026
“The LP30 chip at the heart of the Groq 3 LPX rack carries 512 MB of on-chip SRAM per die, delivering 150 TB/s of memory bandwidth.”
-
E2 GTC 2026 – The Inference Kingdom Expands SemiAnalysis · 24 March 2026
“a significant amount of area taken is up by the 500MB of on-chip SRAM, with a very small amount of area dedicated to MatMul cores that offer 1.2 PFLOPs of FP8 compute”
-
E3 Inside the NVIDIA Vera Rubin Platform: Six New Chips, One AI Supercomputer | NVIDIA Technical Blog NVIDIA · 5 January 2026 Own record
“Up to 288 GB of HBM4 per GPU”
-
E4 How Nvidia's $20 billion Groq 3 LPU deal reshapes the Nvidia Vera Rubin Platform — Samsung 4nm process serves as bedrock for SRAM-based AI accelerator chip Tom's Hardware · 19 March 2026
“That figure dwarfs the 22 TB/s available from the 288 GB of HBM4 on each Rubin GPU.”
Reasoning
An independent report puts the Groq 3 LP30 chip at 512 MB of on-chip SRAM per die [E1], and an independent analyst newsletter describes a chip with 500MB of on-chip SRAM [E2], so "about 500MB" holds. NVIDIA's technical blog lists up to 288 GB of HBM4 per Rubin GPU (E3), and an independent Tom's Hardware report confirms 288 GB of HBM4 on each Rubin GPU (E4). Both figures match the specifications the criteria require.
Cite this claim
Jensen Huang, “The Groq chip has 500MB of SRAM, while a Rubin chip has 288GB of memory.” (said 16 March 2026). Verdict: resolved true. Reliability API, claim c_6142b89e82b9, method 0.1.0+03d23c0e, dataset release v1.1.2. https://epistacklabs.com/claims/c_6142b89e82b9
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
More factual claims by Jensen Huang
-
Mostly true
“we invented a technology called a GPU, and basically a GPU does parallel processing tasks”
-
Resolved false
“Today, NVIDIA’s token cost is the lowest in the world, not by 10%, by X factors, orders of magnitude, all because we did extreme co-design”
-
Mostly true
“It’s the first one to be PCI Express Gen 6.”
-
Mixed
“We have demonstrated that NVFP4 can achieve inference without loss of precision, while delivering significant performance and energy efficiency improvements, and it is equally suitable for training.”
-
Mixed
“Over the past two years, the computational demands for inference have increased approximately 10,000 times, while usage has increased approximately 100 times.”
Verdicts are made automatically; how verdicts are made.