Yep, even if they doubled their production that would still only account for maybe 20-25% market share compared to NVidia if going by the last quarter's GPU shipments of both of them. It may be more if NVidia's shipments were lower, which they certainly are for Q1 at least, but it looks like NVidia's production is back up to normal, so we'll have to see if AMD can increase it's share more than that in the coming months.

I think everyone was kinda' caught off guard on how competitive the RX 9000 series in general was.

Harkens back to the idea that there is no such thing as a "bad GPU" only a "bad price". - AMD just hit the right price even if their Path Tracing  performance and power consumption isn't the best.
Helps their drivers are solid these days as well which is a bonus.

It's a different story here in the UK the 9070XT is still a very bad price basically matching the 5070Ti so if you need a GPU you might as well just buy the 5070Ti as that is objectively the better card at equal price and that keeps getting restocked at or very close to MSRP.

It's interesting to see the different regional choices AMD and Nvidia are making as BofferBraurer pointed out Germany seem to be getting AMD cards whilst the UK are getting plenty of Nvidia restocks.

I think it's not entirely about how competitive AMD got, it's more about how much NVidia stagnated this gen in terms of performance (which allowed AMD to catch up in the first place) plus all it's other blunders like bullshit marketing, bad drivers, blackscreens... pretty much everything AMD had been accused of now suddenly happened to NVidia, which probably had a big effect to shift buying intentions toward AMD this gen.

AI bubble could pop now that cheap new AI models are just copying existing models ("distilling") without the need for extensive training. Nvidia has to serve gamers once again.

There are rumors that demand for Blackwell GPUs for AI is lower than expected, in part for what you've said and also because those chips are too hot and power hungry even for them. That means that Nvidia will have to either negotiate with TSMC to cut orders or use those already booked waffers for gaming GPUs.

We'll see what they'll do, if it's true.

The 4N process nVidia has at TSMC is exclusive, so I would envisage they would keep pumping GPU's out as per their contract and just re-badge them into other markets next year if they have an oversupply.

Once they transition next-years GPU's to a new node, they -should- be able to offload all these GPU's onto the Chinese market as it's not a leading edge product.

One thing NVidia can do is change which chips will be produced. For AI I recon that it would mostly have been the GB 202, and to a lesser degree, the GB 203. The RTX Pro Quadro cards actually only exist with those two chips, even.

But not everybody will buy 5090s, and even the 5080 and 5070Ti will be too much for many to afford. So I expect NVidia to switch production more in favor of the smaller GB 205 and GB 206 for the 5070 and 5060 series of chips instead of the big GB 202 that would just collect dust due to a lack of prospective buyers, at least on the short term.

5060Ti 8GB vs 16GB test from Hardware Unboxed:

Absolutely brutal that even 1080p high is struggling hard in several games on the 8GB version - and even a direct crash to desktop in Indiana Jones  - when the 16GB version is doing close to triple digits at the same time.

NVIDIA GeForce 576.02 driver said to be affected by GPU temperature sensor bug

https://videocardz.com/newz/nvidia-geforce-576-02-driver-said-to-be-affected-by-serious-gpu-temperature-sensor-bug

NVIDIA Rumored To Collaborate With DeepSeek To Develop Custom AI Chips For China; Massive Turnaround Plan For Domestic Markets

https://wccftech.com/nvidia-plans-to-collaborate-with-deepseek-to-develop-custom-ai-chips-for-china/

AMD Believes “Inference” on Mobile Phones and Laptops Is the Future; Claims It Is An Opportunity to Challenge NVIDIA’s AI Dominance

https://wccftech.com/amd-believes-inference-on-mobile-phones-and-laptops-is-the-future/

Intel’s 18A Process Outperforms Intel 3 with Breakthrough Technologies; Superior PPA & Density Scaling Credits To PowerVia

https://wccftech.com/intels-18a-process-outperforms-intel-3-with-breakthrough-technologies/

SPARKLE Refutes Rumors That Suggested Its Working On A 24GB Arc Battlemage GPU

https://wccftech.com/sparkle-refutes-rumors-that-suggested-its-working-on-a-24gb-arc-battlemage-gpu/

Future Intel Arc Graphics could adopt GDDR7 memory, according to job listing

https://videocardz.com/newz/future-intel-arc-graphics-could-adopt-gddr7-memory-according-to-job-listing#disqus_thread

GeForce RTX GPUs gain up to 3-8% synthetic performance with latest 576.02 graphics drivers

https://videocardz.com/newz/geforce-rtx-gpus-gain-up-to-3-8-synthetic-performance-with-latest-572-02-graphics-drivers

Intel Isn’t Done With The Desktop GPU Segment Yet; New Battlemage SKUs Surface Up On Shipping Manifests

https://wccftech.com/intel-isnt-done-with-the-desktop-gpu-segment-yet/

Intel’s next-gen CPU series “Nova Lake-S” to require new LGA-1954 socket

https://videocardz.com/newz/intels-next-gen-cpu-series-nova-lake-s-to-require-new-lga-1954-socket