BofA's Hartnett Says Buy Tech Bonds After 30-Year Low

Bank of America chief investment strategist Michael Hartnett is recommending investors buy US technology bonds, arguing that a borrowing binge by AI hyperscalers has pushed tech debt to levels that offer value after a punishing year. US technology bonds have fallen about 9% in 2026, leaving their total return relative to the broader investment-grade index at a 30-year low. Long-dated bonds from the Magnificent Seven and other AI leaders now yield between 6% and 8%, while CoreWeave's 2031 high-yield bond yields more than 14%.

Context — why this matters now

Hartnett's call puts tech credit in direct contrast with tech equities. Bond investors are being paid more to fund the AI build-out, while shareholders are losing the buyback support that underpinned megacap returns for years. The comparable the report itself gives is stark: hyperscalers issued $182 billion of bonds this year to fund data centres, up from $17 billion in 2024. That roughly tenfold increase in supply is the single largest driver of the underperformance, and it is the reason investors have demanded higher yields to absorb the paper.

The shift in corporate behaviour is the second half of the story. Hyperscalers have raised $89 billion through equity issuance this year, while buybacks have dropped to an annualised $14 billion from $105 billion in 2024. Hartnett said the era of large buybacks is over and that retaining cash benefits bondholders. He pointed to Meta's recent share sale as a sign the company is giving bondholders more consideration, since raising equity rather than adding debt strengthens the balance sheet behind its bonds.

The spending pressure behind all of this is considerable. Bank of America projects AI capital expenditure rising to $1.2 trillion in 2027 from $240 billion in 2024, while free cash flow swings from $240 billion to a negative $180 billion over the same period. Hartnett said bond markets are now imposing discipline on the AI expansion.

The macro backdrop is the obstacle. Hartnett expects investors to stay risk-averse until the global rate-hiking cycle ends, noting central banks have raised rates 42 times in the past six months. Long-dated Treasury yields remain elevated, and that rate backdrop is the main constraint on a broader recovery in risk appetite.

Data — what the numbers show

Credit markets have been signalling stress. S&P Global Market Intelligence data as of Friday showed five-year credit default swap spreads widening sharply across major tech companies since early September. Oracle's spread rose to 276 basis points from 210, and Nvidia's to 129 from 84. Oracle's spread has, however, eased from an October 20 peak of 325, an early sign of stabilisation. S&P Global said in October that investors increasingly treat AI exposure as a separate credit risk.

The before-and-after on supply is the cleanest measure of what changed. AI hyperscalers issued $17 billion of bonds in 2024. This year they have issued $182 billion. Over the same window, buybacks fell from $105 billion in 2024 to an annualised $14 billion, and equity issuance reached $89 billion.

Live market data as of 00:07 UTC today shows Oracle at $130.67, down 9.60% on the session, with an intraday range of $129.27 to $135.20. That equity move sits alongside the credit signal: Oracle's five-year default swap spread has retreated from its October 20 peak of 325 basis points even as the shares sold off. The divergence is the peer comparison that matters here. Credit is pricing stabilisation while equity is pricing something closer to doubt.

For the equity side of Hartnett's framework, the Roundhill Magnificent Seven ETF closed Friday just below his $70 threshold. The VanEck Semiconductor ETF remains about 8% above its own $520 threshold. Hartnett said a year-end rally depends on both funds holding those levels.

Analysis — what it means for markets and sectors

The second-order effect runs through the hyperscaler complex. If Meta's equity raise is the template, then companies funding data centres with stock rather than debt are transferring value from shareholders to creditors. That widens the gap between the two sides of the capital structure, and it explains why Hartnett expects AI bonds' long run of underperformance against AI stocks to end. Bondholders get a stronger balance sheet behind their claim; shareholders absorb the dilution and lose the buyback bid.

Sector exposure follows the capex line. Bank of America's projection of AI capital expenditure rising to $1.2 trillion in 2027 from $240 billion in 2024, against free cash flow swinging from $240 billion to a negative $180 billion, means the funding has to come from somewhere. Debt and equity issuance are the two channels, and both are now open at scale. Utilities, chipmakers and data-centre operators sit downstream of that spending, while the issuers themselves carry the balance-sheet strain.

The counter-argument is that credit markets may be underpricing the risk. Default swap spreads have widened sharply across Big Tech since early September, and S&P Global's October note flagged that investors increasingly treat AI exposure as a separate credit risk. If capex keeps climbing while free cash flow stays negative, the discipline Hartnett describes could arrive as downgrades rather than as voluntary restraint. Oracle's spread easing from its October peak is one data point, not a trend.

Positioning is the tell. Hartnett is steering investors toward the credit side of the AI trade while flagging that equity needs the Magnificent Seven ETF to hold $70. Flow is moving toward the paper that gets paid first, and the semiconductor fund holding about 8% above its threshold suggests the equity bid has not fully broken yet.

Outlook — what to watch next

Hartnett sees relief in the first half of 2027 from geopolitical peace, sharply lower oil prices and an end to rate hikes. Until then, he expects investors to stay risk-averse. The conditionals are specific. A year-end rally depends on the Roundhill Magnificent Seven ETF holding $70 and the VanEck Semiconductor ETF holding $520. The Magnificent Seven fund closed just below its threshold on Friday, while the semiconductor fund remains about 8% above its level.

On the credit side, watch whether Oracle's five-year default swap spread continues to ease from its October 20 peak of 325 basis points, or reverses back toward that level. Nvidia's spread at 129 basis points, up from 84, is the other gauge of whether the widening since early September has run its course. The supply calendar matters too: another year at anything close to $182 billion of hyperscaler issuance would test whether the 6% to 8% yields on long-dated megacap paper are enough compensation.

The rate backdrop remains the gate. Central banks have raised rates 42 times in the past six months, and Hartnett's relief scenario hinges on that cycle ending. Long-dated Treasury yields staying elevated keeps the pressure on both sides of the AI trade.

Frequently Asked Questions

What does Hartnett's tech bond call mean for retail investors?

It reframes where the compensation sits in the AI trade. Retail investors holding megacap equities have absorbed the loss of buyback support, with buybacks falling to an annualised $14 billion from $105 billion in 2024. Bond investors, by contrast, are being paid 6% to 8% on long-dated megacap paper. The trade is about which part of the capital structure gets paid, not about whether AI spending continues.

Why did tech bond returns fall to a 30-year low relative to investment grade?

Supply. Hyperscalers issued $182 billion of bonds this year against $17 billion in 2024, and investors demanded higher yields to absorb the volume. That repricing pushed US technology bonds down about 9% in 2026 and left their total return relative to the broader investment-grade index at a 30-year low. Hartnett's argument is that the supply shock, not deteriorating credit quality, drove the move.