AI Disclosures Add Capacity Limits and Detail Cost Risks
AMD added specific obstacles to customer AI infrastructure buildouts, as XMax and Samsara detailed how adoption and computing costs could affect returns.
The AI accelerator supplier Advanced Micro Devices Inc (AMD) expanded its adoption warnings in August to name insufficient data center capacity, energy availability and construction delays as potential obstacles for customers using its products. In May, AMD said the trajectory of generative AI remained unknown. Its August disclosure retained that uncertainty and specified how infrastructure constraints could delay adoption and impede growth.
AMD also warned that customers may lack capital and request alternative financing or deferred-payment arrangements, and that memory shortages and rising prices may increase data center buildout costs. The connectivity supplier Credo Technology Group Holding Ltd (CRDO) described similar financing exposure in June: some AI customers may be unable to pay for infrastructure or may seek alternative financing and deferred payments. Credo said the semiconductor industry's AI-driven upturn may be unsustainable. Both companies identified customers' ability to fund infrastructure as a risk to their businesses.
The data center and communications product supplier Marvell Technology, Inc. (MRVL) retained its warning from May through August that a significant reduction in AI infrastructure spending “will likely harm our financial results.” Both disclosures said customers could decelerate or reallocate capital expenditure, delaying or reducing product demand. The cloud platform provider CoreWeave, Inc. (CRWV) also maintained its May warning in August: more computing power may cease to produce advances in AI models, and efficiencies from technologies including open-source models may affect demand for its services. These companies continued to question the durability of infrastructure demand through customer spending and technological requirements.
The photonic and electronic semiconductor technology developer Aeluma, Inc. (ALMU) moved from promoting AI data center interconnect opportunities in May to explicitly qualifying its ability to capture AI-related demand in September. Aeluma said it was an early-stage company that had achieved no volume production for any product offering. Growth in AI markets “may not result in meaningful demand” for its specific technologies, citing investment-cycle volatility, competing technologies and customer qualification requirements as commercialization risks.
The AI hardware and software developer XMax Inc. (XMAX) gave its adoption concerns more specific expense and margin context in August. In April, it warned that extended pilots, bespoke integrations and security reviews could lengthen sales cycles and increase upfront costs. In August, XMax said continued sales and marketing spending might fail to produce proportionate revenue increases, and research and development investment might raise near-term expenses without timely benefits. The software platform provider Elastic N.V. (ESTC) likewise warned in June that substantial AI sales and marketing investment might yield no significant directly related revenue “for several years, if at all.”
Elastic also tied its expanding selection of large language models to “compounding uncertainties.” Third-party providers may change outputs, accuracy, capabilities and behavior “without advance notice to us,” creating reliability, customer-trust and compliance risks that Elastic may be unable to anticipate or promptly mitigate. The company separately warned that generative and agentic AI tools could change developer workflows and platform selection, potentially reducing engagement with its products.
The Connected Operations Platform provider Samsara Inc. (IOT) said in September that it was investing significantly in generative AI and AI agents across its solutions and operations. Its warnings linked adoption to flawed algorithms, insufficient or biased datasets and potentially inaccurate recommendations or analyses. Samsara also said increasing use of large language model APIs and related infrastructure exposes it to growing costs.
XMax said it seeks to manage substantial computing costs through improved model efficiency, cloud-provider partnerships and favorable pricing arrangements. Samsara identified a direct risk as usage expands: token-based and consumption-based pricing may affect margins, and AI operating expenses “could increase materially.” The company said there is no assurance that its AI investments will deliver anticipated benefits in a timely or cost-effective manner.