Data center spending remains the dominant theme ahead of first-quarter semiconductor earnings, with major U.S. hyperscale cloud providers preparing to release results in the weeks ahead. Bank of America’s most recent hyperscale capex tracker shows a continuation of that trend, with first-quarter global hyperscale capital expenditure reaching $166 billion - a 13% increase from the prior quarter and an 87% rise compared with the same period a year earlier.
Looking further out, the bank projects hyperscale capital expenditure of $750 billion in 2026 and $872 billion in 2027. Those forward estimates imply year-over-year growth of 57% for 2026 and 16% for 2027. While the firm notes there is uncertainty about whether such elevated spending levels are sustainable through 2027, it adds that recent data center buildout partnerships and announcements in the past two months have improved visibility into the pipeline.
Bank of America also expects cloud hyperscale free cash flow margins to strengthen over the coming years - improving toward a range of 5% to 10% as sales and returns on deployed capital firm up. At the vendor level, the bank highlights a potential wave of AI compute revenue for major semiconductor suppliers. It estimates that AI compute sales from Nvidia, Broadcom, and AMD alone could generate $180 billion to $200 billion of incremental year-over-year sales in 2027, a pace that would exceed 50% growth.
That vendor-level sales projection outstrips the bank’s current projection for hyperscale cloud capital expenditure growth, which it quantifies at $120 billion year-over-year. Bank of America points out that hyperscalers account for only 60% to 70% of total global AI capital expenditure. Based on that share, the firm suggests that total cloud-related spending would likely need to top $1.0 trillion for the vendor revenue expectations to be realized.
Bank of America’s top AI-related stock selections
The bank lists nine companies it views as leading exposures to AI compute and the broader data center buildout, providing price objectives and the valuation multiples underpinning those targets. The selections reflect a mix of compute, networking, optical, and equipment suppliers that participate across chips, ASICs, switches, process equipment, and connectivity.
- Nvidia (NVDA) - Bank of America assigns a $300 price objective based on 28 times 2027 price-to-earnings excluding cash. That valuation captures Nvidia’s leadership in AI compute and networking while also acknowledging risks cited by the bank, including gaming market cyclicality and power access concerns. The note also references a recent legal development in which federal prosecutors charged the co-founder of Super Micro Computer in a scheme to divert servers containing Nvidia chips to Chinese customers in violation of U.S. export controls.
- Broadcom Inc (AVGO) - The firm’s $450 price objective is based on 26 times 2027 price-to-earnings, reflecting expectations for double-digit earnings growth and leading profitability and free cash flow within the semiconductor group. The research also highlights Broadcom’s extended partnership with Google and Anthropic for its TPU technology through 2031. Separately, the bank notes that Seaport Global Securities downgraded Broadcom to Neutral from Buy amid increasing industry constraints.
- Marvell Technology (MRVL) - Bank of America puts a $125 price objective on Marvell using 28 times fiscal 2028 price-to-earnings. The valuation reflects clearer visibility for major customer ASIC projects and a widening AI product footprint spanning connectivity, switching, and compute segments. The bank mentions a strategic partnership between Marvell and Nvidia that included a $2 billion investment from Nvidia, a move that prompted Barclays to upgrade Marvell to Overweight and led to a price target increase from BofA Securities.
- Advanced Micro Devices (AMD) - The bank’s $280 price objective is derived from 27 times 2027 non-GAAP earnings per share, and is premised on AI-driven growth and anticipated CPU share gains. The note references AMD’s own projection of 32% year-over-year revenue growth in the first quarter, which the company ties to rising demand for its data center products. Following those expectations, Erste Group upgraded AMD to Buy from Hold.
- Credo Technology (CRDO) - Credo carries a $160 price objective based on 27 times 2027 price-to-earnings, putting it alongside other high-growth compute and optical semiconductor peers in the bank’s framework. The bank records that Credo reached confidential settlement and license agreements with TE Connectivity and Molex, resolving patent disputes among the companies.
- KLA Corporation (KLAC) - KLA’s $1,850 price objective is based on 40 times 2027 price-to-earnings, a premium multiple the bank justifies with the company’s leading profit margins and extended lead times. At its recent Investor Day, KLA raised its calendar year 2026 outlook and set a 2030 target model that includes revenue of $26 billion. Following that event, Wolfe Research raised its price target on KLA.
- Lam Research (LRCX) - The bank assigns Lam Research a $285 price objective using 41 times 2027 price-to-earnings, reflecting the company’s exposure to ongoing memory and foundry capital spending cycles. Lam announced a leadership change promoting Sesha Varadarajan to Chief Operating Officer. The research also notes that Erste Group downgraded Lam to Hold from Buy, citing elevated margin risks.
- Applied Materials (AMAT) - Applied Materials is given a $420 price objective based on 29 times 2027 price-to-earnings, supported by potential growth in the wafer fabrication equipment market. The bank records that Applied introduced two new chipmaking systems intended for Gate-All-Around transistor production at 2nm and smaller nodes, and that the company will use a new UPS logistics facility in Taiwan as an Asian distribution center.
- M/A-Com (MTSI) - The bank’s $260 price objective on MACOM Technology Solutions is founded on 48 times 2027 price-to-earnings, a multiple the note ties to data center strength and expanding margins. The research cites MACOM’s fiscal first-quarter 2026 results, which posted revenue of $271.6 million and beat expectations; following that report, Benchmark and Stifel raised their price targets on the company.
Market context and implications
The bank’s findings emphasize that vendor revenue potential tied to AI compute could exceed the direct hyperscaler capex trajectory, creating a dependency on broader cloud and enterprise spending for semiconductor vendors to achieve their growth targets. The projection that hyperscalers make up roughly 60% to 70% of global AI capital expenditure underscores the importance of demand outside the largest cloud providers if vendor-level sales are to materialize as modeled.
For investors and market participants, the research frames a cross-section of semiconductor, equipment, and connectivity names as primary beneficiaries of AI infrastructure growth. The price objectives and valuation multiples provided by Bank of America encapsulate both the upside the bank sees and the earnings timeframes it uses to appraise each company.
Conclusion
Bank of America’s tracker and its stock selection list offer a data-driven lens on how hyperscale capex and AI compute demand are shaping investment opportunities across the semiconductor supply chain. While the projections imply sizable vendor revenue upside, the bank is clear that sustaining elevated capital expenditure into 2027 is not guaranteed, and that broader cloud spending would likely need to surpass $1.0 trillion for the most optimistic vendor scenarios to be achieved.