Surging AI Valuations Pressure Software Giants and Cloud Providers Alike

The breakneck rise of generative artificial intelligence is reshaping the technology landscape, unsettling publicly traded software vendors and forcing the world’s largest cloud platforms to commit unprecedented sums to data-center expansion. At the center of the current upheaval is Anthropic, a privately held developer of large language models whose latest financing round reportedly implied a post-money valuation of roughly $380 billion. Market observers speculate that an initial public offering could push that figure to about $500 billion—second only to the near-trillion-dollar valuation many investors now ascribe to OpenAI.

The enthusiasm for these closely held start-ups contrasts sharply with the recent share-price performance of established enterprise-software firms. ServiceNow, Workday, Atlassian and Salesforce have all declined as traders question whether their core products can compete with low-cost AI tools that allow corporate developers to build customized alternatives. Workday even announced the return of co-founder Aneel Bhusri to the chief executive role, a decision viewed by some analysts as a response to doubts about the company’s long-term competitive position in human-resources and finance software.

The cost of staying in the race

While start-ups raise capital with relative ease, the three U.S. hyperscale cloud providers—Amazon Web Services, Microsoft Azure and Google Cloud—have responded by dramatically lifting capital-expenditure targets. Amazon told investors it now plans to spend about $200 billion on infrastructure through 2026, well above the $147 billion Wall Street had modeled. Google offered a similarly higher multi-year outlook, though strong initial reviews for its Gemini model softened the share-price impact. Microsoft, meanwhile, said it has signed up 15 million paid users for Microsoft 365 Copilot out of a potential base exceeding 450 million commercial seats, a penetration rate that investors considered modest.

The resulting cash demands are straining balance sheets traditionally regarded as among the strongest in the sector. Higher spending also limits the capacity for buybacks or dividends and increases reliance on stock-based compensation—issues that have drawn criticism from shareholders focused on near-term returns.

Semiconductor suppliers emerge as short-term winners

Amid the broader technology sell-off, a handful of component vendors have outperformed. Memory-chip producers Micron, SanDisk (Western Digital), Seagate and Western Digital have benefited from a shortage of high-bandwidth memory and large-capacity hard drives used in AI servers. Equipment suppliers Applied Materials, Lam Research and KLA Corp. report full order books as chipmakers rush to add capacity. Thermal-management specialist Vertiv has also logged sizeable bookings for data-center cooling systems, while power-grid suppliers Eaton and GE Vernova are seeing rising demand for switchgear and turbines.

Nvidia, whose graphics processing units are fundamental to most generative-AI deployments, remains pivotal but increasingly faces customer pushback on pricing. Several cloud operators are designing in-house accelerators or partnering with AMD to reduce reliance on Nvidia’s premium hardware, contributing to recent share volatility.

Private-equity exposure and concerns about seat-based pricing

The same skepticism afflicting publicly listed software names is filtering into private markets. Firms such as Thoma Bravo and Vista Equity Partners hold extensive portfolios of enterprise-software companies acquired at valuations predating the current AI enthusiasm. With investors worried that generative tools could reduce overall seat counts—and therefore subscription revenue—appraisals for those assets have fallen. The potential impact remains largely invisible to public-market participants because neither private-equity group is listed.

Imagem: Internet

Another casualty is the assumption that seat-based pricing offers a predictable growth path. Corporate clients increasingly explore AI assistants to automate tasks traditionally handled by junior staff, suggesting future demand for user licenses may plateau or decline. That dynamic complicates forecasting models across the sector.

Questions about real-world adoption

Many chief executives publicly praise AI’s potential, especially for automating repetitive tasks described as “dull, dirty or dangerous.” Privately, however, some concede that current models still generate critical errors, limiting deployment in customer-facing processes. Reliability concerns reinforce the perception that, despite eye-catching valuations, today’s generative platforms remain works in progress. A recent National Institute of Standards and Technology report highlighting model hallucinations underscores those reservations.

Cybersecurity is another area where expectations and capabilities may diverge. Industry specialists question whether emerging AI vendors can match the breach-response track record of established providers such as CrowdStrike, which continues to report customer gains even as software shares broadly retreat.

Outlook

For now, the clear equity-market beneficiaries are companies supplying scarce hardware components and infrastructure services required to train and run large language models. In contrast, legacy software vendors, highly capitalized cloud operators and private-equity portfolios loaded with subscription businesses face mounting skepticism about their long-term positioning. Until visibility improves on actual revenue generated by Anthropic, OpenAI and their peers, investors are likely to scrutinize every additional dollar funnelled into AI data centers, weighing the promise of transformational technology against the immediate cost to corporate cash flow.

Crédito da imagem: CNBC