Wall Street’s Take: AI Valuations Approach Dotcom Heights

This week witnessed a flurry of analyst updates and cautionary notes on AI stock giants. A chorus of voices from major brokerages—including UBS, Truist, BofA, Stifel, and RBC Capital Markets—highlighted both the euphoria and the risks driving the AI investment landscape in 2025. In a reminder of the late 1990s, UBS analysts warned that valuations for top technology and semiconductor stocks are rapidly nearing those seen during the infamous dotcom bubble.

The U.S. tech sector’s HOLT Economic price-to-earnings ratio—a proprietary UBS metric—now sits above 35x. This places parts of the AI segment in the same rarefied air as the post-dotcom peak, raising flags around long-term sustainability. The surge, underpinned by record capital expenditures among hyperscalers (Amazon, Google, Microsoft, Apple, and Meta), has seen big tech’s combined R&D spending in 2024 eclipse the sum of all listed European equities.

Spending is immense: Major U.S. players are set to pour over $350 billion into capital projects in 2025, resulting in tech companies accounting for 37% of projected U.S. economic profit. And yet, many AI use cases remain largely anticipatory rather than revenue-generating in the short term. UBS highlighted a recent MIT study that found 95% of generative AI pilots are not delivering immediate revenue growth—a key indicator that exuberance outpaces fundamentals for now.

OpenAI CEO Sam Altman recently commented that the sector may indeed be in a ‘bubble,’ echoing analyst concerns about near-term cash flow disappointment, especially as power infrastructure and data center limitations threaten to outpace utility capacity.

AMD Upgraded: Traction in AI Data Centers

Amid the market’s bullish fever, Truist Securities upgraded Advanced Micro Devices (AMD) from ‘Hold’ to ‘Buy,’ increasing its price target from $173 to $213. The bullish stance was driven by growing industry sentiment around AMD’s momentum in the data center AI sector. For years, AMD has played the role of challenger to Nvidia’s GPU hegemony, often being viewed as a pricing benchmark more than a true competitor. This sentiment is now evolving.

Truist’s analysts noted that hyperscale customers have begun to see AMD as a true partner, expressing increasing interest in large-scale AMD deployments. This signals a pivotal shift, reminiscent of AMD’s server CPU ascent—from under 1% share in 2018 to roughly 21% just years later as Intel stumbled. While Nvidia’s GPU dominance remains, a sustainable 10% market share for AMD is now seen within reach in the AI data center space. The analysts expect earnings per share for AMD to hit $7.89 by 2027, with its MI355 chip, launched in June 2025, anticipated to drive growth through upcoming quarters.

AMD’s momentum in AI and high-performance computing (HPC) continues to attract market excitement. For Q2 2025, the company reported data center segment revenues exceeding $4.2 billion, marking a 48% year-over-year rise. CEO Lisa Su’s commitment to AI innovation, coupled with new partnerships across cloud hyperscalers like Microsoft Azure and Google Cloud, provide further upside catalysts as demand for GPU alternatives intensifies.

Bubble Watch: AI Leaders at Risk?

While AI remains a dominant corporate narrative—one in every four U.S. company earnings calls now mentions AI, according to UBS—the downside risks are mounting. The HOLT model indicates that much of the sector’s value depends on optimistic future cash flows, leaving little margin for error. Analysts are particularly wary of overextension in the ‘Mag 7’ stocks—Apple, Nvidia, Amazon, Alphabet, Microsoft, Meta, and Broadcom—which, collectively, have lifted U.S. market indices but may be increasingly vulnerable to any cash flow shocks or regulatory headwinds.

UBS and other strategists are now advising investors to rebalance away from U.S.-centric AI stocks, highlighting opportunities in non-U.S. quality growth names, global ETFs with secular growth track records outside of U.S. technology, and sectors such as utilities, energy, and real estate, all of which show low correlation to tech performance.

Marvell Downgraded: Softer AI Growth Prospects

Bank of America shifted its view on Marvell Technology Inc (MRVL), downgrading the chipmaker from ‘Buy’ to ‘Neutral’ and lowering its price target to $78 from $90. This move was triggered by weaker-than-expected Q2 2025 results and rising uncertainty regarding Marvell’s AI pipeline, particularly its role in Microsoft’s Maia initiative and Amazon’s nascent 3nm chip program. Marvell’s projected data center segment growth for 2026 was slashed, with BofA’s analysts citing ‘incrementally higher uncertainty’ and a $400 million drag on FY27 sales projections.

Still, Marvell’s core IP portfolio and $2.5 billion in auto unit sale proceeds provide a financial safety net—and potential powder for share buybacks or future acquisitions. The stock, trading at about 20x-21x PE after the earnings miss, may have valuation support, but investors are advised to monitor ongoing developments closely as the AI hardware market grows more competitive.

Ambiq Micro: Edge AI IPO Gains and Risk Factors

Fresh off its IPO, Ambiq Micro (AMBQ) has drawn mixed reactions from analysts. Stifel initiated coverage with a ‘Buy’ and a $45 target price, praising Ambiq’s lower-power “SPOT” architecture as a competitive advantage in the rapidly expanding Edge AI market. With more than 270 million chips shipped and established relationships with giants like Google and Garmin, Ambiq is set for broader customer adoption, particularly through its upcoming Atomiq platform.

However, others remain cautious. UBS and Bank of America launched coverage at ‘Neutral’ citing execution risks and high customer concentration. Both firms note profitability is not expected until 2028, with the company reliant on achieving gross margin expansion to 54%. Ambiq’s post-IPO volatility has moderated, and its $657 million market cap will be put to the test as it scales beyond wearables into wider IoT and edge computing use cases—a sector forecast by IDC to reach more than $69 billion by 2026.

Software’s Resilience: “AI Won’t Kill Software,” Says RBC

RBC Capital Markets countered the rising narrative that AI will render traditional software obsolete. While software stocks excluding mega-caps have underperformed (the IGV software ETF would show double-digit declines without Microsoft, Oracle, and Palantir), RBC’s analysts maintain that incumbents’ data, customer relationships, and distribution channels will enable them to thrive through the AI transition.

The bank sees a spectrum between extreme disruption and incremental innovation, forecasting that AI will benefit select incumbents (such as Microsoft, Intuit, HubSpot, MongoDB, and Pegasystems) while stimulating new entrants via M&A and technological advances. However, broad AI monetization—on a scale that fundamentally transforms software P&Ls—isn’t expected until 2028 or later. In the interim, indirect benefits may show up as increased platform usage, retention, or higher deal win rates.

Recent pullbacks in names like Dynatrace, HubSpot, MongoDB, ServiceNow, and Snowflake could offer value for long-term investors willing to weather volatility and ride the wave of enterprise AI adoption.

Summary: Opportunity amid Caution

AI remains the central story for investors and enterprises alike in 2025, but as Wall Street’s sharp analyst shifts indicate, risk and reward are tightly intertwined. While companies like AMD and Ambiq are emerging as beneficiaries of changing market dynamics, lofty valuations and execution hurdles are forcing investors to weigh upside against historical precedent—and reminding all that, in the world of AI, cycles of hype and consolidation are likely to remain in force for years to come.