BofA Hikes 2026 Semiconductor Forecast to $1.3 Trillion, Eyes $2T by 2030

Wall Street’s most closely watched semiconductor analyst just handed the chip industry a remarkable new milestone to aim for. In a research note published April 13, Bank of America’s Vivek Arya raised his 2026 global semiconductor revenue forecast to $1.3 trillion—a $300 billion upward revision from the estimate he issued just four months ago, and a figure that would represent a 30% year-over-year surge from 2025 levels.

The revision is not a minor trim. It is the equivalent of adding an entirely new semiconductor industry the size of 2021’s total market onto the existing 2025 base—in a single year.

What Drove the Upgrade

Arya’s thesis is straightforward and AI-centric. He projects a 43% year-over-year jump in compute and storage semiconductors, driven by hyperscaler capital expenditure programs that have accelerated well beyond earlier projections. Data center buildout, AI training clusters, inference infrastructure, and the supporting networking and memory stack are collectively consuming silicon at a pace that earlier industry models consistently underestimated.

The upgrade reflects a recognition that AI workloads are fundamentally different from previous demand drivers. Unlike consumer electronics cycles—which plateau at smartphone upgrade saturation—AI infrastructure spending is driven by the economics of model training and inference at scale: each new model generation requires more compute than the last, and deployment at commercial scale requires orders of magnitude more inference capacity. There is no natural ceiling visible from current vantage points.

Arya’s top six picks for 2026 were selected specifically for dominant market shares typically in the 70–75% range in their respective segments:

• Nvidia (NVDA) — AI GPU dominance in training and inference
• Broadcom (AVGO) — Custom AI ASICs and networking silicon
• Marvell (MRVL) — Networking and custom silicon for hyperscalers
• AMD (AMD) — Data center GPU second-source and CPU
• Lam Research (LRCX) — Etch equipment critical for advanced node production
• KLA (KLAC) — Semiconductor process control and inspection

The inclusion of equipment makers Lam Research and KLA is notable. It signals that BofA’s bull case is not confined to fabless AI chip designers but extends to the picks-and-shovels layer—the companies whose tools are required to manufacture chips at any node, regardless of which architectures ultimately win in the market.

Consumer Weakness Persists

Not all segments share in the optimism. Arya modeled only a 9% decline in wireless communications chip revenue—a relatively mild correction given how much smartphone demand has softened—but warned that consumer demand could remain sluggish into 2027. That projection weighs most heavily on companies like Qualcomm and Skyworks that derive the bulk of their revenue from smartphone RF and application processors.

The bifurcation between AI-driven enterprise silicon and consumer-facing chips is arguably the defining market dynamic of 2025–2027. Investors who treated “semiconductors” as a monolithic category in prior cycles are increasingly having to distinguish between companies with AI infrastructure exposure versus those tethered to consumer refresh cycles.

The Cloud Capex Warning

BofA’s bull case comes with a significant embedded caveat. For chip vendors to hit their 2027 sales targets, Arya’s model requires global cloud capital expenditure to exceed $1 trillion—meaningfully above the current consensus estimate of $872 billion. That gap is not trivial.

The “capex gap” is the most common pushback bulls receive on the semiconductor thesis. Hyperscalers—Microsoft, Amazon, Google, and Meta—have signaled record capex commitments for 2026, and earnings call language has been consistently optimistic about continued AI infrastructure spending. But the question of whether that spending rate can be maintained through 2027, especially in the face of potential macroeconomic headwinds, trade policy uncertainty, and tariff exposure on semiconductor supply chains, remains genuinely open.

Arya’s note does not dismiss these risks but argues that the structural AI demand driver is powerful enough to sustain above-consensus growth even if individual quarters show volatility.

The Road to $2 Trillion

The longer arc of BofA’s projection is even more striking. The bank now forecasts the total semiconductor market reaching $2 trillion by 2030, implying a roughly 20% compound annual growth rate over the next four years. For context, the industry took approximately 50 years to reach its first $500 billion year—and is now being projected to double that in less than a decade.

The $2 trillion scenario assumes continued AI model scaling, the emergence of physical AI (robotics, autonomous vehicles, edge inference), and the gradual electrification of everything from factory floors to residential energy management creating new silicon demand categories.

Industry Context: From $1T to $1.3T in One Quarter

The velocity of the estimate revisions is almost as striking as the figures themselves. BofA’s prior forecast was already bullish by historical standards; the $300 billion upward revision in a single quarter reflects how much the AI infrastructure build has accelerated beyond even optimistic scenarios. Global chip sales data for the first months of 2026 have consistently exceeded analyst expectations, with monthly figures running 60–70% above year-ago levels.

The semiconductor industry—long accustomed to boom-bust inventory cycles driven by consumer electronics—is experiencing a structurally different demand curve. AI infrastructure is inherently long-cycle capital investment. Hyperscalers plan data center builds on multi-year horizons; chip supply chains respond accordingly. That means the volatility profile of semiconductor stocks may itself be changing, as annualized revenue forecasts become more predictable and less subject to sudden inventory correction.

Whether $1.3 trillion materializes or the year ends closer to $1.1 trillion, one fact seems beyond dispute: the silicon century is well and truly underway.