Microsoft is seeking to raise at least $25 billion through a US investment-grade bond offering in up to 10 tranches, with maturities from two to 40 years. JPMorgan, UBS and Wells Fargo are managing the sale, with settlement expected Nov. 4, 2026. Initial price guidance was around 50–55 basis points over Treasuries for two-year fixed-rate notes and 130–135 basis points for 40-year bonds due 2066. Proceeds are for general corporate purposes, including capital expenditures, working capital, buybacks, acquisitions and debt repayment.
Microsoft is returning to the bond market after a long absence as large tech companies borrow heavily to fund AI infrastructure. Microsoft has historically relied on cash flow, but AI data-center and chip spending is pushing even top-rated firms toward debt markets. Microsoft is one of only two publicly traded US companies with top-tier AAA ratings from Moody's and S&P, alongside Johnson & Johnson.
Microsoft is seeking at least $25 billion in a US investment-grade bond sale, its first since 2017. Its last high-grade debt sale was $17 billion in January 2017, after a roughly $19.75 billion offering in 2016 to help fund the LinkedIn acquisition.
TechCrunch reported that Mintive has raised funding from Cowboy Ventures, Coatue, and angels including Jeff Dean and Andrej Karpathy.
TechCrunch reported that Wabi raised $55 million at a $1.25 billion valuation and now describes itself as a “personal agent.”
TechCrunch reported that Bluesky blocked links to a site with leaked DHS employee data amid government takedown requests.
TechCrunch reported that DoorDash is acquiring Instacart for $17 billion, combining restaurant and grocery delivery.
ETwenty-five billion dollars. One market account says Microsoft may be about to borrow that much for the first time in almost a decade.
STwenty-five billion? Microsoft prints cash. Why borrow?
EBecause even a company sitting on this much cash has to fund AI data centers somehow. The account Unusual Whales posted this morning that Microsoft is looking to raise at least twenty-five billion in its first bond sale since twenty seventeen.
SI remember that seventeen billion dollar deal. It was right after the LinkedIn acquisition, right?
EYeah, January twenty seventeen. The LinkedIn purchase was almost twenty billion, and they borrowed to help pay for it.
SSo this is different. This is about AI infrastructure.
EMostly, but not entirely. The filing says proceeds go to general corporate purposes, including capital spending, working capital, share buybacks, acquisitions, and paying down debt.
SSo basically the corporate equivalent of 'I'll figure it out later.'
EMore like 'we'll take the cheap money now.' Their CFO Amy Hood told investors they'd borrow when it's attractive to preserve cash and flexibility. Apparently now is attractive.
SHere's my concern. Microsoft is one of only two publicly traded US companies with the top triple-A rating from both Moody's and S&P. The other one is Johnson and Johnson. If Microsoft has to tap the bond market to fund AI, what does that say about how expensive this buildout is?
EIt says debt is cheap and they'd rather not drain their balance sheet. If you have a triple-A rating and you don't use it, what's the point?
SThe point is discipline. When everyone is borrowing to build AI data centers at the same time, I start to wonder if we're seeing a capex bubble.
ENo, no, wait. In 1999, companies borrowed to fund dot-com hype with no revenue. Microsoft has real cloud revenue, real enterprise customers, and a real backlog. It's not the same story.
SBut AI demand is still a bet. Those data centers are only worth what customers actually pay for the compute. If enterprise adoption slows, that's a lot of expensive real estate.
EThat's exactly why long-term debt makes sense. You're matching a forty-year asset with forty-year financing. If they funded all of this with short-term cash flow, they'd have to cut elsewhere if there's a hiccup.
SAlright, the forty-year piece is interesting. What are the terms?
EInitial price talk is around fifty to fifty-five basis points over Treasuries for two-year notes, and one hundred thirty to one hundred thirty-five for forty-year bonds due twenty sixty-six.
SThat's cheap money, I'll give you that. But it's also a sign of how much capital AI is sucking up. Alphabet priced twenty-five billion in August. Now Microsoft. This is becoming routine.
EAnd that routine is exactly what makes it less scary. These are the only companies with balance sheets strong enough to borrow this much without flinching. The AI race is expensive, but the biggest players can afford it.
SI'm not saying Microsoft is in trouble. I'm saying we should watch whether the revenue from AI catches up to the spending. If it doesn't, these bonds will look like a very long bet.
EFair enough. But that's the nature of infrastructure investment. You build before demand peaks, not after.
SOkay, I'll concede the capital structure point. It's smart to borrow cheap long-term debt. But I still think the AI buildout has real execution risk.
EAnd I think the risk of not building is far greater for Microsoft. If they sit out, they lose their cloud lead.
SThat's actually my worry. They're forced to spend because competitors are. That's not always a healthy dynamic.
ECompetition is what drives innovation. I'd rather see them spend than get complacent.
SSpoken like someone who doesn't have to sign the check.
ENot yet. Give it time.
EDifferent world entirely, but connected by capital: the FCC may have just made a big move against Chinese labs.
SWhat now?
EOne market account reports the FCC voted to ban all Chinese labs from testing US electronics starting December twenty twenty-eight.
SI can't confirm that vote from any official source yet, so treat it as a claim. But if true, that's a significant escalation.
EThe rationale would be pretty clear: keep Chinese government-linked labs away from American tech standards and testing.
SSure, but US electronics companies rely on those labs for certification and testing. Cutting them out entirely could cause delays and higher costs.
EMaybe, but national security concerns around Chinese access to US electronics are real. If they're testing devices before they hit the market, they see everything.
SThat's the argument. But I'd like to see the actual text of the rule before believing a market account.
ESame. We'll see if an official announcement lands. If it does, it's a big deal for the chip and hardware sector.
SAnd for every company that sells electronics in the US. December twenty twenty-eight is a few years away, but supply chains move slowly.
EThey'd need to start planning now, assuming it's real.
SWhich is why I'm not dismissing it. But unverified is unverified.
EExactly. We'll keep an eye on it.