Nvidia apparently takes place too quickly with AI, as with the end of the GTC 2025, we now know that the ultra range of Blackwell, with Vera Rubin, should be deposited on the markets at a “concerning” rate.
NVIDIA’s generations of Blackwell IA are six months apart with regard to volume production; An attempt to give no space to its competitors
Well, we are all quite aware of the fact that Nvidia, in its current domination on the AI market, is simply due to the way the company has managed to capitalize on the media threshing of the “at first” AI “, and not only that, but with the roadmap of the company’s product, the competitors failed to make up for it, and the group of bands AI continued to run with Nvidia in the driver. According to an analysis of the famous analyst DAN NYSTEDTThe current cadence of the Green team could simply exhaust the supply chain, and we have already seen some signs of the faults present in Blackwell GB200.
In May 2024, at Computex, Nvidia announced that it would accelerate its AI roadmap, repelling new architectures with a difference of one year, to catch up with market expectations and ensure that competitors do not obtain their “respiratory space”. After the announcement, we saw the liberation “forced” “forced” of the Blackwell GB 20000 servers from NVIDIA in the fourth quarter of 2024, which would be in “limited quantities”, only to ensure that the markets remain confident with Nvidia as a leader, but did that pay?
We saw Blackwell confronted with a massive rate of return that interrupted the supply chain, and the CEO of Nvidia, Jensen Huang, admitted the fault of architecture. The problems were not resolved until the beginning of the first quarter of 2025, with servers manufacturers like Foxconn Ramping Up Production by the end of the first quarter 2025. After the supply chain finally obtained a certain catch on the Blackwell AI range, Nvidia announced the Blackwell Ultra GB300 range, which is said to the production of H2 2025, which means that the annual category Now reduced, which is certainly something to note.
This may be what Nvidia plans. Dominate IA computer markets by forcing industry to need more recent architectures and, ultimately, not allowing competitors to present themselves to their solutions. From the AMD Instinct MI300 range, we have seen Nvidia release nearly three new queues, if you include the Topper generation. It is therefore sure to say that Team Green is expanding the AI supply chain, or the other probability is that this is what the company has wanted from the start.
To top it off, Vera Rubin, the architecture announced at the GTC 2025 and will lower on the market at the end of 2026, would be published six months earlier. This is due to the fact that SK Hynix is already planning to produce in mass HBM4 by Q3-Q4 of 2025, which means that Nvidia can potentially lower Rubin on the market by T1 2026 or a “small” launch by the end of this year. Memory manufacturers cannot wait for their HBM4 to be integrated into a delay after production, and the only company to announce the use of the standard is NVIDIA for the moment.
It will be interesting to see how the NVIDIA AI roadmap is revealed by the end of the year. As Jensen says, “the more you buy, the more you save.” Whether it turns out to be true or not is something for the future.
