• TD Cowen reiterated its Buy rating and $290 price target on AMD, citing strong AI spending trends and the upcoming ramp of the Helios AI platform.
  • The firm views the recent stock pullback as an attractive entry point, anticipating a 'noisy but promising' 2026 as Helios gains traction.
  • HPE's commitment to commercialize the Helios rack-scale system worldwide in 2026 marks a critical validation and go-to-market milestone for AMD.

TD Cowen has reaffirmed its confidence in Advanced Micro Devices Inc., maintaining a Buy rating and a $290 price target. In a note to clients, the firm highlighted that the recent weakness in AMD's share price presents a compelling buying opportunity, with the company's strategic positioning in artificial intelligence infrastructure set to strengthen significantly through 2026.

The core of this bullish outlook hinges on the commercialization of AMD's Helios rack-scale AI architecture. The platform, designed to operate a full rack of accelerators as a single system, represents AMD's most direct challenge yet to Nvidia's dominance in high-performance AI clusters. According to people familiar with the matter, the technical specifications are formidable: each Helios rack integrates 72 of AMD's next-generation Instinct MI455X GPUs, delivering up to 2.9 exaFLOPS of FP4 performance and boasting 31 terabytes of HBM4 memory.

A major step toward turning these specs into revenue came on December 2, 2025, when Hewlett Packard Enterprise announced it would offer the Helios system worldwide starting in 2026. This partnership with a leading OEM provides AMD with a crucial commercialization pathway and immediate distribution reach. "HPE becoming a launch partner is a significant validator," said one industry analyst who asked not to be named. "It signals to the market that this isn't just a prototype; it's a product with serious backing."

The architecture itself, built on Meta's Open Compute Project (OCP) specifications, is a strategic differentiator. By leveraging open standards, AMD is appealing to hyperscalers and large cloud providers—like OpenAI, which has shown interest—who are actively seeking to diversify their AI hardware suppliers and avoid vendor lock-in. This open approach is intended to encourage a broader ecosystem of partners to develop around the platform.

TD Cowen's characterization of 2026 as "noisy but promising" acknowledges the competitive and execution hurdles ahead. Ramping a system of this complexity, which also incorporates AMD's EPYC Venice CPUs and Pensando Vulcano NICs, involves significant integration challenges. Furthermore, AMD must convince customers to commit to large-scale deployments against Nvidia's deeply entrenched ecosystem and proven track record.

AMD did not immediately respond to a request for additional comment on the Helios production timeline. However, the HPE announcement and analyst backing suggest the pieces are falling into place for a serious competitive push in the AI infrastructure race, with the real financial impact beginning to materialize next year.

Correction: An earlier version of this article misstated the performance figure for FP8. The Helios platform delivers up to 1.4 exaFLOPS of FP8 performance.