January 7, 2026, 07:31 PST
- Nvidia says its next-generation Vera Rubin platform is in “full production” as it lines up cloud and AI customers
- AMD rolled out new MI455 and MI440X chips and previewed a 2027 MI500 series as it presses its OpenAI ties
- Analysts’ targets still favor Nvidia and AMD, while Intel draws mixed calls despite a fresh upgrade
Nvidia’s next-generation Vera Rubin platform is now in “full production,” CEO Jensen Huang said at CES in Las Vegas, giving investors a sharper sense of when the hardware driving the AI boom is due to land.
Why it matters now: CES has turned into a pit stop on the chip industry’s road map. Cloud firms and big AI developers show up looking for the next wave of data center gear, kicking the tires ahead of earnings season. But the stakes keep rising as companies move from training giant models to running them at scale for consumers and businesses.
Analysts are tracking those launches closely, trying to figure out who comes out on top in 2026. Nvidia and AMD are still viewed as the cleanest bets on demand for graphics processing units (GPUs) — chips used to train and run AI models — and “inference,” the step where a trained model generates answers. Intel has also crept back onto the radar, though investors don’t appear fully sold.
Huang said Rubin’s flagship server will carry 72 GPUs and 36 central processing units (CPUs), and can be linked into “pods” holding more than 1,000 Rubin chips. He said the platform could boost the efficiency of generating “tokens” — the chunks of text AI systems process — by 10 times, and add “context memory storage” so chatbots can handle long prompts faster. https://www.reuters.com/world/china/nvidia…
AMD used CES to unveil its MI455 AI processors aimed at server racks bought by customers such as ChatGPT maker OpenAI. It also rolled out MI440X, pitched at companies that prefer running AI on their own systems rather than on purpose-built clusters. CEO Lisa Su pointed to its “Helios” rack-scale design as a blueprint for “yotta-scale” computing, saying a single rack can deliver up to 3 AI exaflops — a measure of computing power. https://www.reuters.com/business/amd-unvei… https://www.amd.com/en/newsroom/press-rele…
On Wall Street, Bernstein analyst Stacy Rasgon said Rubin was “in full production, and looks like a monster,” TipRanks reported. TipRanks also flagged Piper Sandler analyst Harsh Kumar, who reiterated a Buy on AMD with a $280 price target. Based on TipRanks data, Nvidia’s average price target suggested roughly 40% upside, compared with about 32% for AMD, while Intel’s consensus target pointed to a slight downside. https://www.tipranks.com/news/amd-vs-nvda-…
In a Zacks commentary carried on Nasdaq, Nvidia has seen the sharper upward revisions to earnings estimates, while AMD’s projections have been mostly flat in recent weeks. Zacks projected AMD earnings of $3.96 a share for FY2025 and $6.26 for FY2026. For Nvidia, it penciled in $4.66 for FY2026 and $7.24 for FY2027. AMD is set to report on Feb. 3, and Nvidia is due on Feb. 25.
Intel, meanwhile, unveiled Core Ultra Series 3 laptop processors made on its 18A manufacturing process, and said pre-orders would start Jan. 6, with systems shipping worldwide from Jan. 27. Melius Research analyst Ben Reitzes upgraded Intel to Buy and lifted his price target to $50, MarketWatch reported, arguing the foundry and packaging story could end up mattering more in 2026 than investors expect.
