The artificial intelligence hardware landscape is showing early signs of realignment, as analysts at Citrini Research point to growing traction for Google’s tensor processing units, or TPUs.
While Nvidia remains the dominant supplier of graphics processing units — the engine of today’s AI boom — the success of Google’s Gemini 3 model has prompted renewed scrutiny of whether alternative architectures could gain meaningful ground.
Gemini 3’s performance, fuelled by TPUs rather than GPUs, is prompting investors and developers to reassess long-term dependency on Nvidia’s technology.
With Google positioning itself as both a model developer and chip supplier, the TPU ecosystem now appears to be moving from experimental to increasingly commercial.
“While the writing has been on the wall for some time, the market’s perception of Google has dramatically reversed over the past several months – transforming from an AI loser, bleeding its search dominance, to a stalking horse destined to undercut the most consensus AI winners..,” analysts at Citrini wrote.
Meta interest underscores early commercial shift
Reports earlier this week suggested Meta Platforms may be in talks with Google to purchase or lease billions of dollars’ worth of TPUs.
While no agreement has been confirmed, the development signals growing interest in training frontier models without relying exclusively on Nvidia hardware.
Citrini notes that the TPU opportunity has been building for months.
Alphabet’s share price has outpaced Nvidia’s recently, reflecting evolving perceptions of the two companies’ positions in the AI race.
For years Nvidia’s CUDA software ecosystem has been a formidable moat — one that locked developers into GPU-based acceleration.
But partnership discussions with Meta, Anthropic and Fluidstack suggest that moat could narrow over time.
Suppliers see upside in a more diversified AI chip market
Citrini Research argues that the relevant question is no longer whether TPUs are competitive, but how fast the market could widen if hyperscalers diversify beyond Nvidia.
If TPU adoption accelerates, it could reroute demand toward companies embedded in that supply chain.
Citrini has mapped out a group of firms that could benefit directly from broader TPU deployment, naming publicly traded players including TSMC (TSM), ASE (TW:3711), Amkor (AMKR), Micron (MU), Lumentum (LITE), TTM Technologies (TTMI), SK Hynix (KR:000660), Unimicron (TW:3037), Apple (AAPL), SiTime (SITM) and Macom (MTSI).
Semiconductor manufacturers TSMC, ASE and Amkor are expected to gain from advanced packaging requirements, particularly CoWoS assembly, which remains crucial for training-scale AI chips.
Memory suppliers including Micron, SK Hynix and Samsung may also benefit, as high-bandwidth memory remains a bottleneck for both TPU- and GPU-driven computing.
Printed circuit board suppliers such as TTM Technologies, Unimicron and Isu Petasys are seen as well-placed to serve next-generation TPU configurations, which require higher density designs and specialised routing.
Optical component makers, including Lumentum, could see improving demand through their role in optical circuit switching for data-centre-scale workloads.
Citrini also highlights potential adopters such as Apple, which may use TPUs for cloud-based inference in future Siri Intelligence deployments.
Chipmakers SiTime and MACOM are likewise positioned to benefit from a shift toward 1.6 terabit optical modules for high-speed data transmission.
TPU adoption still faces high switching costs
Despite growing momentum, analysts warn that transitioning away from Nvidia architecture remains expensive and technically complex.
Software compatibility, developer familiarity and deployment logistics continue to favour GPUs.
Nvidia still commands close to 90% share of the AI accelerator market, and any fragmentation would likely occur gradually rather than overnight.
Even so, Citrini Research sees the risk to Nvidia’s margins rising over the long term.
If TPUs gain commercial adoption at scale, NVIDIA’s premium pricing — underpinned by historically tight supply — could come under pressure.
For now, Nvidia remains ahead — but the terrain beneath the AI hardware market is shifting.
Whether Google’s TPU ecosystem evolves into a genuine second pillar of global AI infrastructure may determine the next phase of growth across the semiconductor supply chain.
The post Google’s TPU momentum could create new winners across the AI supply chain: these stocks could benefit appeared first on Invezz