GPU Software Development Engineer

Intel's Razor Lake-AX could deliver nearly 4x the AI power of Panther Lake. With 700 TOPS of AI compute on the roadmap, Intel's Razor Lake-AX could bring a major performance boost to next-generation robotics and AI systems. VIEW GALLERY - 4 Senior Tech Reporter Published Oct 18, 2026 3:25 PM CDT 1 minute & 30 seconds read time Intel plans to launch Razor Lake-AX in 2028 with up to 700 TOPS, nearly 3.9 times Panther Lake's rated AI throughput. The processor targets robotics and physical AI, while rumored GPU and memory details-and potential laptop and compact-PC use-remain unconfirmed. Its specifications, power use, price, and real-world performance are unknown. Voice: Hassam Nasir 0:00 / 2:54 Intel plans to introduce its Razor Lake-AX processor in 2028, targeting up to 700 TOPS of AI computing performance. The chip is part of the company's push into robotics and physical AI and could deliver a substantial increase over its current Panther Lake platform. Intel executive Cho Min-sung outlined the company's robotics roadmap at the Semiconductors Powering Humanoids conference in South Korea on October 16, according to TechPowerUp. The roadmap places Nova Lake-based robotics solutions in early 2027, followed by Razor Lake-AX in 2028, as manufacturers prepare to scale production of autonomous mobile robots and humanoid systems. VIEW GALLERY - 4 IMAGES The 700 TOPS target represents a substantial jump over Panther Lake, which Intel rates at up to 180 platform TOPS across its processing engines. On paper, Razor Lake-AX would offer nearly 3.9 times that level of theoretical AI throughput. Although the company has yet to reveal a complete specification sheet, earlier leaks suggest Razor Lake-AX could feature a powerful integrated GPU. We recently covered rumors of 16-core and 32-core Xe3P graphics configurations, alongside the possibility of on-package LPDDR5X or LPDDR6 memory. The flagship GPU tile reportedly measures 162.84 mm^2 (nearly triple the size of Panther Lake's 12-core Arc B390). Another report suggested the processor might reuse Nova Lake's Coyote Cove performance cores and Arctic Wolf efficiency cores, rather than adopt new CPU architectures. If these specifications materialize, the processor could also appeal to high-end laptops and compact PCs requiring graphics and AI capability without a dedicated GPU. Those consumer-oriented configurations, however, remain speculative. Intel's growing focus on robotics extends beyond traditional AI PCs, with its existing Panther Lake platform already supporting local perception and control through dedicated hardware and software tools. Story Q&A. Powered by TweakBot From TweakTown's coverage of this story. What AI workloads could Intel's Razor Lake-AX run locally in humanoid robots and autonomous mobile robots? Razor Lake-AX could run more complex AI workloads directly on humanoid robots and autonomous mobile robots, with less reliance on cloud processing. Intel's existing Panther Lake platform already supports local perception and control, but no specific Razor Lake-AX workloads have been identified. How might Razor Lake-AX's projected AI throughput change the robotics workloads Intel's Panther Lake platform can handle locally? What Intel hardware and software support enables local perception and control on Panther Lake, and could it be relevant to Razor Lake-AX? How could Razor Lake-AX's rumored Xe3P graphics configurations support graphics-heavy robotics workloads? Want something else? Ask TweakBot. Razor Lake-AX could take those capabilities further, enabling more complex AI workloads to run directly on robotic systems with less reliance on cloud processing. However, despite its ambitious 700 TOPS target, Intel has not yet revealed key details, including power consumption, final specifications, pricing, and real-world performance.

Intel chief executive Lip-Bu Tan says the Trump administration has no policy to use its minority stake to influence the chipmaker's strategy. "They don't have any influence on the governance of the company," he told Nikkei in Tokyo. The US government converted federal funding into an $8.9 billion investment last year, giving it a 9.9% stake, later diluted to around 8.2%. The stake is passive, with no board seat. Tan said government backing helps Intel make the long-term investments a competitive foundry business needs. He said Intel is "80% done" with its restructuring, and named AI-related chip packaging and application-specific integrated circuits as other growth areas. Asked about China's chip industry, Tan said: "Today, I think it's not that far behind.

Intel shares fell 3.94% to $91.61 at 11:17 ET on October 16 after TSMC said it can support customers if rival advanced-packaging capacity falls short. The comment raises the stakes for Intel's expansion, though nothing shows Intel has lost a customer or order. Intel's Foundry unit posted $5.765 billion in second-quarter revenue and a $2.089 billion operating loss, about 36.2% of segment revenue. The figures cover the whole Foundry business, not packaging alone, and include internal transactions. The share price stands 182.92% above GuruFocus's $32.38 GF Value estimate. Investors appear to be paying for a turnaround that still requires packaging milestones, production at scale, and narrower losses.