Products You May Like
Nvidia sent shockwaves through the semiconductor industry this week when it announced explosive quarterly earnings. The company is experiencing a meteoric rise in revenues and market cap thanks to a seemingly limitless demand for its AI hardware, which has caused long wait times for its GPUs. According to the company, those wait times will not improve when its next-generation hardware arrives, as on the earnings call, both Nvidia’s CFO and CEO said they expect demand for its next-generation “Blackwell” data center hardware to far outstrip supply.
According to a transcript of the earnings call via SeekingAlpha, the company’s executives discussed the previous quarter and the future. Nvidia is expected to unveil Blackwell at GTC 2024, which takes place in early March. The conference used to be about supercomputing in general but has now been labeled an “AI conference,” to nobody’s surprise. On the call, Nvidia’s CFO Colette Kress stated, “We expect our next-generation products to be supply constrained as demand far exceeds supply,” which is unsurprising. Kress also said the supply is improving for its existing H100 products, leading one analyst to ask about the discrepancy.
Nvidia’s roadmap shows “Ada Lovelace” next in 2025, but we will have to wait and see if that comes to fruition.
Credit: Nvidia
An analyst named Stacy Rasgon queried, “Why does that get constrained as Hopper is easing up?” At this point, Jensen jumped in, saying this is just how it goes with new products from Nvidia these days. He stated that even though the company’s supply chain is doing great work, a new product has to ramp from zero to a “very large number” quickly, and thus it’s impossible to have enough hardware initially. He stated, “And so whenever we have a new generation of products—and right now, we are ramping H200’s. There is no way we can reasonably keep up on demand in the short term as we ramp.”
The elephant in the room here is whether or not this same imbalance of supply and demand will apply to the company’s next-generation GeForce products. We all remember the pandemic and the recent insanity involving RTX 4090 GPUs, and nobody wants to see that situation again soon. However, given Nvidia’s dominance in AI and gaming, scarcity seems almost unavoidable when there’s a new product launch, at least on the high end.
Regardless, Nvidia’s AI GPUs are now the hottest commodity since Beanie Babies, so we have no doubt they will be hard to come by for the foreseeable future. One interesting twist is that Blackwell is expected to be Nvidia’s first data center GPU with a chiplet design, as noted by Tom’s Hardware. That decision will help improve yields, though the advanced packaging required could also burden TSMC, hindering supply. The only bright spot in the pronouncement is that it could make its current H100 Hopper GPUs easier to purchase, as wait times for that GPU are still several months, according to most reports. However, that’s down from up to one year previously.