Patrick Moorhead

Microsoft and Oracle last reported more than $1.3 trillion of signed contracts they haven't delivered yet. Both say demand is outrunning supply. The AI question isn't demand. It's how fast they can build. Thursday I'm on @YahooFinance right after $INTC reports. It's The Week Ahead In Technology for October 26, 2026. $MSFT, $GOOGL and $AMZN report back to back. I care less about beats than backlog becoming revenue. @Oracle hosts AI World and its investor day. It expects about 13% of its backlog to become revenue within a year; Microsoft, about 30%. $INTC: the 14A customer-commitment window is open. Customers set the clock. $KLAC, $SKHY and $NOW report. ServiceNow: agents in booked revenue, not pilots. I'm in New York for @cloudera EVOLVE26 to talk sovereign AI. So who turns backlog into revenue fastest?

“For you” tab is shite again.

“The more you buy the more you save”. Sound familiar? $NVDA

Most software devs I know are PC gamers. Does Sam really want to alienate this segment?

I do try and smile on camera!! It was a busy week.

Here's what I said on @theinformation's TITV with @akashparichha on $INTC, currently down ~29% at $65.53: 1. Is this all out of Intel's hands? Substrates are short and Windows PCs are weak; Microsoft guided Windows OEM down mid-30s. It's an industry issue, but I'll hold final judgment until AMD reports. 2. OK with no free cash flow in 2027? Yes. Foundry capex is a requirement. Data center margin is 41% vs 23% a year ago; foundry loses 28 cents per revenue dollar, similar to hyperscalers going FCF negative to build capacity. AMD doesn't have to deal with this. 3. When does foundry make money? 2029 or 2030 (my belief, not what company is saying). The customer packaging announcements Lip-Bu said are coming in the next two quarters matter. 4. Who are the customers? Likely Google and AWS on packaging. On wafers, I believe Nvidia and Apple come first: they need a second source and need more US manufacturing and capacity. 5. Read-through to $AMD? Data center is full steam ahead, and TSMC will do AMD favors. PC demand isn't gone: PCs don't live 7 to 10 years.

Here's what I said on @SquawkStreet with @SaraEisen and @carlquintanilla on $INTC, down ~28% at $66.16: 1. Does the reaction make sense? It really doesn't. People bought Intel in the $120s for foundry, and nothing changed there, though execution questions are fair. Everyone knew PCs would be way down; Intel could have flagged it sooner. 2. Does PC weakness slow foundry and AI? In 2027, server and PC chips align on 18A and 18A-P, and that's where Intel makes a lot more money. PC demand doesn't go away. Memory relief looks like 2028, not 2027. 3. $AMD has handily beaten Intel over six months. Risk in AMD's quarter? Less. Both face the same PC inventory issues, but TSMC is going out of its way to supply AMD wafers. It's $TSMC's foundry vs Intel's. 4. How long is the turnaround? Lip-Bu never promised a couple of years. It's multiyear, with some execution hiccups. If $AMD shows the same substrate pain next week, Intel's problems are the market's. I don't think it will. @intel $INTC

Headed on @CNBC at 11 EST at the top of the show to break down the $INTC quarter, the guide, and long-term prognosis.

$INTC off the call. 10 things not in the release that were meaningful: 1. Five-year view: Intel sees server CPU TAM up to 5x, $40B to $200B+. Agents, agents, agents. 2. This year and next: server CPU units, all vendors, up 30%+ each year, with shortages into 2028. 3. Q4: data center up mid-teens QoQ, client down 20%+ on inventory cuts. 4. Israel fab outage: ~$300M of Q4 revenue, ~2 points of gross margin, nearly all 10nm server chips. Q1 revenue hit similar. 5. PC weakness frees wafers: Oregon's Panther Lake 18A capacity shifts to Intel 3 Xeon, 2H27. 6. PC units -12% in 2026, -10% in 2027; no memory relief seen in 2027. 7. Custom chips: +50% QoQ, 2x+ YoY, below-average margins. 8. Intel is open to fabbing CPUs on any ISA, Arm included, and offering modular x86. 9. 2027 investment year: capex up strongly, FCF probably not positive. 10. A couple of packaging customers named in the next two quarters; 14A 1.0 PDK by Q1. @intel

Here’s what I said on @YahooFinance on in response asked to five questions on $INTC : 1. What stands out? A really good quarter: met on revenue, beat on almost everything else. It's the guide. Q4 is ~$1.5B light on weaker PCs and scarce substrates, and those hit everybody. Could be issues at their Israel fan. 2. Memory prices? PCs are down 15% to 25% on the shortage and memory prices that doubled or tripled. Demand isn't destroyed: nobody rides a five-year-old PC another five years. 3. AI demand for CPUs? Data center grew 78%. That's AI, and it lasts. Agentic AI needs one CPU per accelerator, sometimes two. 4. Foundry deals? Intel won't pre-announce them. @LipBuTan1 said he'd only spike capex with a real customer, and 2027 capex is going up. That tells me there are big wafer and packaging customers. 5. What stops the bleeding? Nothing fixes the PC market soon. I want Intel's plan to get enough substrates, and a straight read from Lip-Bu on foundry wafers and packaging. @intel

Good $INTC quarter, very challenging guide. We will have to wait until the call to get the details. Has to be from one or all of these: 1. weakening PC (see $MSFT comments) 2. ⁠scarcity of substrates 3. ⁠incident at Intel’s fab in Israel

Headed on @YahooFinance to discuss the $INTC quarter and forecast. I will be on in 15 minutes.

Wait, are we supposed to all be using Sonnet now???

$MSFT, day two: 16 target raises, no cuts, and down ~1.3% at midday anyway. -DA Davidson $550 to $650 -Citizens $550 to $625 -Stifel $575 to $650 -Goldman Sachs $640 to $700 -Bernstein $662 to $722 -Morgan Stanley $600 to $650 -Wells Fargo $725 to $750 Also raised: Piper Sandler, TD Cowen and Scotiabank to $650; Truist, KeyBanc, StoneX and Cantor to $630; Barclays to $610; Guggenheim to $595. Equal-weighted average of all 16: $610 to $651, up 6.8%. My math. Bernstein has the right frame: the efficiency that carried Azure this quarter shows up in margins later. Same logic behind my October call that Maia is a margin lever.

$NOW, day two: down ~25.5% at $111.72, extending the selloff on the Q4 cRPO guide. 19 re-rates, zero rating changes. What an overreaction. Equal-weighted, the 19 went to $162 from $164, down 1.5%. The 10 cuts went to $157 from $171, down 8%. My math. The street moved a fraction of the stock's 25%. Cuts: -Truist $180 to $150 -BTIG $170 to $150 -Morgan Stanley $180 to $160 -Wells Fargo $190 to $175 -Citi $170 to $157 -RBC $176 to $165 -UBS $150 to $140 (Neutral) -Barclays $160 to $152 -Mizuho $170 to $165 -TD Cowen $165 to $160 Raises: -Baird $125 to $185 -KeyBanc $85 to $110 (Underweight) -Jefferies $140 to $145

$ORCL is down ~3% at ~$140 after yesterday's ~11% jump on @Oracle's investor day. What moved it yesterday was a $20B cap on new capital through FY30, half equity, half debt, and free cash flow positive in 2H FY28, $30B+ in FY29. FY30 targets held: $225B+ revenue, $21+ non-GAAP EPS, same as last October. Since June I've said demand isn't the debate, the funding model is. Oracle answered it. Some thoughts: -On a 1 GW deal, customer-supplied hardware cuts Oracle's upfront capital to ~$6B from ~$37B. Less revenue, higher margins and returns. -Build speed is the new debate, as I said Sunday. ~7 GW more by end of FY28 means averaging above fiscal Q1's ~850 MW a quarter. My math. -My colleague @LeoneDataAI nailed it at AI World: agents on Oracle's apps create its own OCI demand, not just GPUs rented to labs. Live Fusion agents: 1,000 in May, 3,500 in September. What I'd still like to see is a quarter above 850 MW delivered. The December print is the first test. Nice job Oracle.

This is good. And apparently a built in automated token router, too, to optimize costs.

Ironically, almost to the same day 14 years ago, Microsoft introduced the Surface RT with Windows and capacitive touch. Some of my favorite Apple exec quotes against touch on Macs: Tim Cook, CEO, 2012: •April earnings call, asked about tablet and PC convergence: "you can converge a toaster and a refrigerator," but the tradeoffs leave a product nobody likes. •To the WSJ, on a touch iMac: "this kind of reach for me isn't a terribly intuitive thing." He wouldn't discuss touch notebooks, calling that a roadmap question. •October call, on Microsoft's Surface: you could design "a car that flies and floats," but it wouldn't do those things well. •Craig Federighi, SVP Software Engineering, Jan. 2014 (Macworld): He said it's easy to "slap a touchscreen on a piece of hardware," then asked whether that's a good experience. "We believe, no." •Federighi, Oct. 2014 (CNET): "We don't think it's the right interface, honestly." He said constantly reaching forward from a desk to touch the screen is uncomfortable. Apple was "not all that interested in building one." •Jony Ive, Chief Design Officer, Oct. 2016 (CNET): Apple ruled out a touchscreen Mac years earlier because "it wasn't particularly useful or an appropriate application of multitouch." •Phil Schiller, SVP Worldwide Marketing, Oct. and Nov. 2016 (Touch Bar launch): ◦To CNET: engineering was better spent on the "best keyboard-trackpad experience possible." ◦To The Independent: Apple had looked at it, and "it's a bad experience." ◦To Backchannel: a touch desktop would "become absurd." "You can't optimize for both. It's the lowest common denominator thinking." •Federighi, Oct. 2016 (CNET): Apple had working prototypes years earlier. But grafting touch onto a system built around a precise pointer "really compromises the experience." •Federighi, June 2018 (Wired): He said he's "not into touchscreens." He said lifting your arm to poke at a screen is "a pretty fatiguing thing to do." Good to see Apple changed their mind.

Installed.