Categories
News

Nvidia Has Been the Undisputed King of the Artificial Intelligence (AI) Revolution. Has the Chipmaker Finally Met Its Match?


There is not any query that Nvidia (NASDAQ: NVDA) has been the principal beneficiary of current advances in synthetic intelligence (AI). The corporate’s graphics processing units (GPUs) shortly grew to become the gold customary for generative AI, capturing a shocking 92% of the information heart GPU market, in keeping with market researcher IoT Analytics. Nvidia has parlayed that dominance into 5 consecutive quarters of triple-digit year-over-year gross sales and revenue development.

Many opponents have tried to maintain up with the firm’s relentless tempo of innovation, however none have succeeded. Simply this yr, Nvidia revised its product launch cadence from two years to yearly, making it even more durable for rivals to compete.

Begin Your Mornings Smarter! Get up with Breakfast information in your inbox each market day. Sign Up For Free »

Nevertheless, a current entrant into the AI market is making waves and will mark the first actual competitors Nvidia has confronted.

A person pushing a virtual AI button, surrounded by various technology icons.
Picture supply: Getty Photographs.

Cerebras Techniques is an AI firm based in 2016, and there have just lately been rumblings of an IPO on the horizon. The corporate believes that “AI is the most transformative expertise of our era.”

Cerebras developed the Wafer-Scale Engine (WSE) — a big semiconductor that’s taking a distinct method to accelerating AI. The WSE boasts 4 trillion transistors and integrates 900,000 compute cores and 44 gigabytes of Static Random Entry Reminiscence (SRAM) into the chip itself.

Cerebras claims that its distinctive building reduces latency — or the lag ensuing from information transmission — making the third-generation WSE “the world’s quickest commercially obtainable AI coaching and inference answer.” In August, Cerebras launched what it known as “the world’s quickest AI inference,” which it claims is 20 occasions quicker than Nvidia’s GPU-based options at a fraction of the value.

In a press launch that dropped final week, Cerebras up to date its claims, saying it tripled its “industry-leading inference efficiency, setting [a] new all-time file.” The corporate mentioned its checks with Llama 3.2 — the just lately upgraded generative AI mannequin from Meta Platforms — have been “16x quicker than any recognized GPU answer, and 68x quicker than hyperscale clouds.”

Whereas AI-centric efforts by Nvidia and Cerebras have some overlap, it is essential to take a step again and put the rivalry in context.

Nvidia’s chips have a observe file courting again 25 years and have stood the take a look at of time. These GPUs dominate a range of duties and markets, together with online game graphics playing cards, information facilities, earlier branches of AI, and — most just lately — generative AI.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *