Analyst: NVIDIA has 'effectively canceled' B100 AI GPU over design flaw: B200A to replace it

NVIDIA's upcoming earnings report is ruffling some feathers already, one analyst saying Blackwell AI GPUs design flaws are seeing the B100 AI GPU "effectively canceled".

The design flaws plaguing NVIDIA's new Blackwell AI GPUs hit headlines a couple of weeks ago, where we began hearing about design flaws that analyst firm KeyBanc says NVIDIA will need to "respin" the Blackwell tile that will cause a 3-month delay on shipments.

KeyBanc explained: "Given the Blackwell delay, we believe NVIDIA will prioritize the ramp of B200 for hyperscalers and has effectively canceled B100, which will be replaced with a lower cost/performance GPU (B200A) targeted at enterprise customers".

That's pretty big news right there: an analyst says that NVIDIA has effectively canceled its new B100 AI GPU -- because of glaring flaws that can't see it shipped in its state -- and will push out a revised B200A just like we reported a few weeks ago.

• Read more: NVIDIA hits major roadblocks with Blackwell AI GPU: revised B200A AI GPU
• Read more: NVIDIA's new Blackwell AI GPUs 'encountering major issues' which require redesign, big delays
• Read more: NVIDIA's next-gen Blackwell AI GPUs delayed, rumor has it 'design flaws' are to blame

2

Now, what are the differences between B100 and B200A?

B100:

• Packaging: CoWoS-L
• Memory: up to 192GB HBM3E
• Logic Dies: 2
• Power: 700W

B200A:

• Packaging: CoWoS-S
• Memory: up to 144GB HBM3E
• Logic Dies: 1
• Power: 700W

At the time -- August 4 to be precise -- SemiAnalysis reported that B200 AI GPUs were meant to be shipping in Q4 2024 but issues with the Blackwell die needing a possible redesign, or the bridge dies inside TSMC's new CoWoS-L advanced packaging needs a redesign. The B200A is born from that, and is expected in the second half of 2025.

The second half of 2025, when they were meant to be shipping in Q4 2024 is a big delay, which is why NVIDIA is pushing its B200 out at its limits in Q4 2024 in extremely limited supplies before more B200 chips are made in 2025.

NVIDIA effectively canceling its B100 AI GPU is an interesting thing to see happen in a post on X by an analyst, but it's a weird clown world we live in during these times.

The full post on X reads: Supply chain feedback is indicating performance issues, and a resulting chip respin with the Blackwell tile is resulting in a one-quarter delay but is not expected to have any impact to near-term results and guidance. We believe modest expectations for Blackwell shipments in FQ3 have been backfilled with higher Hopper bookings. We expect NVDA to report beat/raise results, in which upside will be driven by strong demand for Hopper GPUs.

Given the Blackwell delay, we believe NVDA will prioritize the ramp of B200 for hyperscalers and has effectively canceled B100, which will be replaced with a lower cost/performance GPU (B200A) targeted at enterprise customers. The ramp of GB200 has also been impacted, but we are still expecting modest volumes of NVL36/72 to ship in F4Q. Despite the delay, our FY26 rev/EPS est of $222B/$5.16 remain unchanged, which compares to cons est of $168B/ $3.79. We see favorable risk/reward into earnings with NVDA trading at 25x our FY26 EPS. Reiterate OW.