Nasdaq 100 Falls as AI Trade Shows New Cracks, Guests Weigh Microsoft's Bond Sale and Alphabet's Miss

The discussion opened with Microsoft's offering of more than $25 billion in Triple-A rated bonds, which the host described as the day's headline credit story. Gil Luria, Head of Technology Research at D.A. Davidson, and Robert Schiffman, Senior Tech Credit Analyst at Bloomberg Intelligence, were brought in to discuss technology from equity and credit perspectives.

Luria said Amazon, Google and Meta have now moved to negative cash flow. Microsoft, by contrast, still generates positive cash flow even though it plans to put about $200 billion into capital expenditure. In his words, "Microsoft's business is so strong that in spite of the fact that they'll put 200 billion of CapEx into the ground, they still have positive cash flow, so they don't actually need the capital." He added that this is typically the best time to raise capital because it secures the best price.

Luria said his model runs out to 2028, with revenue of roughly $469 billion and gross profit of about $0.65 per dollar of revenue. He compared that with a margin of about four times the $0.37 per dollar he cited for gross margin, and called Microsoft "a profit machine."

Asked how embedded Microsoft's profit base is, Luria pointed to the broad use of its products. Most companies use Microsoft Office and Windows, which together cover the vast majority of the market. On top of that, Microsoft sells other software infrastructure, cloud storage and, through Azure, now sells compute to most companies. "They're incredibly entrenched in the global enterprise market. And that's what makes them so bulletproof," Luria said, adding that this is what supports their margins and makes their revenue sticky.

Luria said the good news for Microsoft was that AI compute accelerated Azure to 48% growth. He stressed that this is not a startup but a business of about $20 billion that grew 48%. Describing his own "amateur take" on the figures, he said Azure was the key factor, and that the change in its growth rate was the key determinant of his read.

Schiffman said Microsoft's offering, at least $25 billion and its first US investment-grade sale in about nine years, is an example of why this space is exciting: a borrower rated Triple-A, higher than the US government, is coming to the market. He read it as an implication that Microsoft is about to spend much more than it has been.

Schiffman contrasted this with peers that must be more conservative with their capital structures and are issuing equity, citing a roughly $100 billion equity issuance and Alphabet issuing about $85 billion of equity. He said Microsoft is still buying back stock. He agreed that Luria is correct that Microsoft generates free cash flow in theory, but argued that it is not left over after shareholder returns, and that Microsoft will keep returning cash to shareholders while raising capital expenditure.

Schiffman noted that yields are rising and that a previous guest had suggested five-year yields could go above 6%. Even so, he argued that the weighted average cost of capital remains low relative to the cost of equity. He said, "These hyperscalers are cost agnostic and quite frankly, relative to the cost of equity, it's still so low." He expects more borrowing from Microsoft, and suggested the 5% figure could move toward 10%, if not 15%.

Schiffman recalled that when Nvidia gave its revenue guidance a few years ago, his first thought was that AI could be an existential threat to Google search, and suggested the previous night might have been that turning point. Luria agreed, saying everyone's initial reaction when ChatGPT appeared was that search was at risk because people would start their queries in a chat. He said Google search held up for two or three years and that the market became complacent. "And then last night there was a very abrupt deceleration," he said, with search growth falling from 19% two quarters earlier to 17% a quarter earlier and then to 12%, and expected to fall even lower next quarter.

Alphabet fell a record 13% on a sales miss. Its YouTube business and subscription line also came in below expectations. The panel asked whether this was an inflection point, possibly a shift toward generative commerce, with people using chats to shop for insurance or flights. The logic offered was that when a search happens through a chat, no one clicks through an advertisement, and that a bot searching on a user's behalf means an ad cannot be sold. The panel suggested this might be why the reaction to Google's results has been so strong, and that this quarter may be when the shift showed up.

The host said the question about the depreciation trend of chips had gone unanswered by others and passed it to the guests. Luria said the figure is a moving target. A year or two ago, the expectation was that Nvidia chips would depreciate over three or four years, not the six years discussed today. Data centers are now selling capacity on chips that are eight or nine years old, so the pendulum is swinging toward the view that these chips have longer economic lives.

Luria's view was that the truth is probably somewhere in the middle. Demand is so vast that chips are being used for longer, but over the longer term he expects a return to a useful life of about five to six years.

Schiffman argued that companies are putting their money where their mouths are through residual value guarantees and large debt deals. He cited a $60 billion Anthropic deal with Broadcom and a $40 billion chip deal involving SpaceX and Nvidia, which he said implies that these chips will be worth much more than the market has been implying. "I think we've overshot on how quickly they're depreciating," he said, noting chips that are ten years old are still in service, so he is not so worried.

Luria said that from a credit perspective he does not care whether Alphabet is worth $3 trillion or $4 trillion. "I think it's a wild overreaction. This is a $4 trillion market cap company," he said, adding that people treat one print as the death of a business when it takes many cuts. He said he understands the fears about the death of search, but that while it affects equity, he does not think it will affect spreads or Alphabet's ability to borrow. He pointed out that Alphabet can borrow 30-year bonds within 10 to 15 basis points of Microsoft, which he said provides enormous support for the spending it wants to do.

Asked whether cloud revenue and profitability could offset weakness in core search advertising, Luria said the cloud business could become a larger part of the story, but that it is unlike Amazon Web Services, which now makes up most of Amazon's profit and value. Google's cloud business is still closer to 20% of profit, so it will take a while to offset the deceleration in digital advertising. He said the growth was spectacular, with the business roughly doubling from a $20 billion base. Google is also now selling its TPU chips, which he described as the only viable alternative to Nvidia chips, beyond giving capacity within its own cloud.

Asked for his single best buy, Luria named Micron. He said everyone is still underestimating how much memory is needed to keep building AI capacity, because models run better and faster with more memory and have better context windows. He added that Micron still trades at about six times earnings. Schiffman was asked to respond, and said memory would keep doing well.

Schiffman pointed to Oracle, whose shares rose sharply the previous day. He said the conversation is moving from how much capital expenditure is being spent and how much needs to be borrowed to when these companies will turn free cash flow positive. Oracle said it would turn free-cash-flow positive and become self-financing, which Schiffman said flipped a switch for it. Its credit default swaps tightened by about 50 basis points and its long bonds tightened by about 50 basis points over the past day.

Schiffman expects the trend to continue, and that within the next year the hyperscalers will use the same language. He suggested that in two years, the question will be why they are not buying back more stock and raising dividends.