NVIDIA Warns of Limited Blackwell GPU Supply Amid Surging AI Demand

NVIDIA "ironically" anticipates it will run out of Blackwell B100 GPUs before their respective releases, given the company's high demand.

NVIDIA's Next-Gen Blackwell AI GPUs to Face Supply Constraints, CFO Predicts

Whether they are potential clients or investors, the next-generation AI accelerators from Team Green have energized everyone, as it is expected that the Blackwell lineup will revolutionize computing performance and client demand. Given these circumstances, NVIDIA's CFO Colette Kress has predicted that a comparable predicament may arise with Blackwell compared to the Hopper H100.

"We are delighted that the supply of Hopper architecture products is improving. Demand for Hopper remains very strong. We expect our next-generation products to be supply constrained as demand far exceeds supply," NVIDIA's CFO Colette Kress said.

Everyone anticipates a limited supply, given that the Blackwell AI GPUs are a relatively new product line; thus, a restricted supply is expected. Due to the immense anticipation surrounding the forthcoming accelerators from the AI industry, NVIDIA may experience a disruption in the supply and demand chain, which could result in inflated pricing, lengthy order backlogs, and numerous other issues for the B100s. This is customary in the sector, accompanied by each additional product launch. Considering that Blackwell AI GPUs are designed for the "explosive" AI market, we anticipate they will succeed tremendously.

To provide a brief overview of the Blackwell B100, NVIDIA intends to incorporate state-of-the-art HBM3e memory and implement a chipset design expected to yield substantial performance improvements. The GPUs are anticipated to be introduced by the 18th of March at GTC 2024, where an announcement will most likely be made.

NVIDIA Sets Stage for GTC 2024: AI and High-Performance Computing in Spotlight​

From the previous report of Wccftech last year, the schedule for NVIDIA's GTC 2024 (Graphics Technology Conference), which will be held at the San Jose Conventional Center beginning on March 18, 2024, has been disclosed.

Although GTC 2024 is still quite some time away, Team Green has nearly five months prematurely announced the official dates. The conference is highly regarded within the technology community, primarily because prominent companies such as NVIDIA reveal details regarding their forthcoming products slated for the annual "upgrade cycle." Not to mention that NVIDIA CEO Jensen Huang frequently makes an aesthetically pleasing appearance, as his demeanor establishes the ambiance for the remainder of the occasion.

Although NVIDIA has not disclosed specifics regarding the content of its keynote address, it will center on artificial intelligence. NVIDIA has consistently bragged about its "Blackwell" GPUs for the high-performance computing (HPC) and artificial intelligence (AI) sectors, asserting that they will indeed usher in the "future" of AI computing processors.

Although information regarding the architecture is limited, the GPUs are anticipated to employ TSMC's 3nm process. Furthermore, they are rumored to be the initial iteration of NVIDIA's AI GPUs to incorporate a chipset design that would supplant the conventional "monolithic" standard. Blackwell GPUs are anticipated to supplant the most coveted AI chips currently available, NVIDIA Hopper, with production beginning in the latter half of 2024.

Consumers are reportedly oblivious to any developments concerning the company, at least regarding industry rumors. The gaming-oriented "Ada Lovelace-Next" GPUs from NVIDIA are anticipated to debut in 2025; however, based on the past few GTCs, it is improbable that we will learn anything about them.

Except for a few consumer-grade software announcements (SteerLM/Nemo demonstrations at previous GTCs), they are primarily focused on HPC and AI. Everyone is looking forward to GTC 2024, and we are curious how the announcements will influence the industry's future.

Photo: BoliviaInteligente/Unsplash