Hightower: AI capex is structural

Hightower Advisors began 2026 with a question that is never far from a client's mind: is the AI trade a bubble? In its January 6 Weekly Wisdom commentary, a special edition devoted to the year ahead, the firm says no. It builds the case on a single figure: $360 billion, the expected 2025 AI-related capital expenditure across the major hyperscalers. That money reflects multiyear commitments to data centers, computing infrastructure, and model development.

Financing is the first defense. In the second half of 2025, Amazon raised $15 billion in debt. Meta raised $30 billion. Alphabet raised $24 billion. Hightower emphasizes these raises were backed by cash flow, not leverage. Amazon produced $130.7 billion in operating cash flow over the past year. Alphabet produced $73.9 billion in free cash flow. Meta produced $44.9 billion. The bond market has effectively ratified the thesis. Amazon's debt trades at a 53.0 basis point spread over Treasuries. Meta's sits at 71.4 basis points. Alphabet's is 51.4 basis points. Those are investment-grade spreads, a sign the market trusts the cash flows behind the borrowing. For an advisor running a fixed-income sleeve, the credit takeaway is simple: the market treats AI buildout debt as high-grade risk.

The most client-ready number in the report is the adoption gap. Hightower reports that 98% of Fortune 500 companies have experimented with generative AI. Only 26% have deployed it at scale. Consumer-driven usage accounts for 85% of AI activity. Together, the figures describe an enterprise market still getting ready to put the technology to work. The infrastructure is built, but the enterprise deployment is roughly three-quarters empty.

Cybersecurity is the second theme, and it follows directly from the first. The report argues that AI-driven threats are accelerating, especially from non-human agents, while AI-assisted coding expands the attack surface. Hightower points to Robinhood, Coinbase, Microsoft, and Google as companies where AI tools now write a large share of new code. At some of those companies, the share reaches 30% to more than half. It also cites the November 2025 mass cyberattack involving Anthropic as an example of the urgency. For a client portfolio, cybersecurity is the closest thing to a second-derivative trade: it monetizes the AI buildout without betting on any single AI winner.

Hightower calls the material 'a mix of familiar and newer developments' and positions it as a framework rather than a forecast. In a portfolio review, that is the right way to use it. The capex figure speaks to the size of the buildout. The bond spreads say whether it is financed. The adoption gap says how much room remains. The cybersecurity theme points to who gets paid next. Each answer maps to a holding or exposure the client already owns.

One risk deserves attention. The 85% consumer-driven usage figure has a flip side. It supports the growth case now, but if enterprise deployment stalls, the same figure would argue the other way, leaving the AI story dependent on consumer discretionary spending. Hightower reads the gap as a reason for more growth. A careful advisor will show the client both interpretations.

The firm frames these as longer-term shifts it expects 'in the months ahead.' That gives advisors room to talk about structural trends rather than quarter-to-quarter forecasts. AI capex and cybersecurity have been portfolio topics since well before 2026. The novelty here is the measurement: a handful of concrete numbers can turn a vague 'what about AI?' into a review of capex, cash flow, adoption rates, and security spending. If the 26% deployment figure starts to climb, the conversation will shift from whether the AI trade is real to which companies are best positioned to turn it into profit.