BofA: $1.5T AI Buildout Makes Chip Dip a Buy

What did Bank of America say about the chip stock selloff?

Bank of America analyst Vivek Arya called the current chip stock pullback a "summer reset" — not a structural break in AI demand. In a July 6 report, Arya noted the PHLX Semiconductor Index had dropped 11% since the start of Q3, after surging 88% in Q2. He said this matches historical seasonal weakness for the period and expects a rebound in the fall.

"History suggests periods of consolidation are often followed by renewed momentum as investors regain confidence in the next leg of earnings and capex growth," Arya said, according to Benzinga.

How big is the AI infrastructure buildout?

Global cloud and AI infrastructure spending is projected to approach $1.5 trillion by 2027, up 40% to 50% from current levels. Arya attributed this growth to persistent demand for compute, accelerated AI agent deployment, and structural supply-side constraints.

This scale of AI capex spending is the foundation of Arya's thesis. He argues that hyperscalers remain focused on maximizing AI utilization — not cutting infrastructure budgets — which keeps chip demand robust.

Which 7 stocks does BofA expect to lead?

As 2027 spending visibility improves in the second half of 2026, Arya expects market leadership to rotate back to companies tied directly to AI capital expenditures. Here are the seven names he highlighted:

You might also like

Grok 4.5 Launches July 9 as xAI's Opus-Class

Rapidus Raises ¥424.95B for 2nm Chip Push

OpenAI Launches GPT-Live-1 Voice Model

Pope Leo XIV Issues AI Encyclical at UN Summit

Why is Micron BofA's top pick?

Arya singled out Micron as one of the market's biggest AI mispricings. Memory now accounts for roughly 35% to 40% of AI cloud capital spending — two to three times higher than historical levels. Yet memory stocks still trade at around 10x forward P/E, which Arya considers severely undervalued.

The reason for the discount: investors fear memory pricing will revert to its traditional boom-and-bust cycle. Bank of America disagrees. Memory is moving "from a cyclical commodity to a strategic AI enabler," Arya said, per Bitget's coverage of the report.

Arya believes long-term supply agreements between memory suppliers and hyperscale customers are making pricing more durable and revenue more predictable. That structural shift, he argues, justifies higher valuation multiples over time.

Bank of America reiterated its Buy rating on Micron and kept a $1,550 price target, implying roughly 59% upside from current levels.

Does the rise of Chinese AI models threaten chip demand?

Arya addressed this directly. Chinese open-weight models — including DeepSeek, Kimi, Qwen, and GLM — have rapidly closed the gap with US frontier labs. As of July 4 third-party benchmark rankings, US models from Anthropic and OpenAI still lead, but Chinese models hold 8 of the top 16 spots. The highest-ranked Chinese model is GLM 5.2 from Zhipu (Z.ai), an open-weight model with 750 billion parameters and a one-million-token context window.

Arya's view: cheaper AI models pressure software profit margins, but they expand AI adoption. More deployment means more demand for compute, memory, networking, and power infrastructure. "The bigger risk is to model economics, not semiconductor demand," the report stated.

This dynamic connects to broader trends we track at iCharles — Meta's cloud AI buildout and OpenAI's revenue targets both assume inference costs keep falling while infrastructure demand keeps climbing.

Arya also noted that Nvidia is actively participating in open-source community building. This extends Nvidia's hardware ecosystem reach to small and mid-scale AI adopters who lack direct access to frontier labs.

What does the seasonal pattern suggest?

The PHLX Semiconductor Index — tracked by the iShares Semiconductor ETF (SOXX) — historically shows weakness in Q3. The 11% pullback after an 88% Q2 surge fits that pattern. Arya's call is that the dip is temporary, and that leadership will rotate back to AI capex-linked names as 2027 spending plans become clearer in H2 2026.

Here's what we know so far: the report's core argument rests on the idea that AI infrastructure demand is structurally different from prior chip cycles — and that memory, in particular, is being mispriced by a market still applying old commodity frameworks.

For builders and founders watching AI infrastructure spending trends, the BofA thesis points to one concrete milestone: improved visibility into 2027 hyperscaler capex plans, expected to emerge in the second half of 2026.