Intel's 9% jump prices a mid-2028 wafer promise at 92 times this year's earnings

Only one of the three stories carrying Intel's price has already turned into billings, and it is the data centre and AI line, up 59% year on year ^[15]; the other two are a wafer supply agreement whose first products arrive no earlier than mid-2028 ^[8] and a price increase that Taiwan's DigiTimes, citing supply chain sources, describes as under consideration for October rather than announced ^[12].

The foundry arithmetic is where the re-rating gets interesting, or rather where it gets uncomfortable: Intel Foundry turned over $5.8bn last quarter and lost $2.1bn at the operating line, and external customers accounted for $293mn of that revenue ^[11], which is about 5% of the segment ^[1] and means the operating loss ran roughly seven times what outsiders actually paid ^[2]. Terafab is the trade against that ratio. Northland's estimate has the project making 22.4 million advanced logic wafers a year at full tilt against about 8 million of TSMC advanced capacity today ^[9], call it 2.8 times ^[9], though the same note concedes that not all of Terafab's output would be made in Intel plants ^[10], so the number describes a pool Intel can bid into rather than an order book it holds. Intel says it has run more than a million wafers through ASML's High NA EUV tools ^[14], which demonstrates process readiness rather than customer demand.

The price leg is the one that leaves the building. A 10% list increase on PC CPUs during a shortage falls almost entirely to gross margin, because the wafer cost behind it does not move, and RBC's Srini Pajjuri has x86 server CPU volumes rising 15% to 20% to meet demand ^[13], which is the same shortage seen from the buyer's side. Whoever signs PC and server orders next cycle funds this turnaround before any foundry customer does.

Then the multiple. At $104.47 against expected earnings of $1.14 a share the stock trades on about 92 times this year ^[4], and the $150 scenario implies about 132 ^[17]. On the 2027 number of $2.04, Mizuho's $92 is roughly 45 times and Northland's $120 roughly 59 ^[5], a $28 gap worth about 27% of the current price ^[6]; the more telling figure is that the same 2027 estimate was $1.51 ninety days ago ^[18], a 35% revision ^[7]. A market capitalisation of $552.1bn implies about $6bn of net income on this year's EPS ^[10]. And the price is still about 26% below the June 22 high of $140.94 ^[5][8], so the tape has already run one 30% drawdown through this story.

Pricing power is the only leg that shows up in the next two prints, and it is also the leg with the shortest life, because shortages end when capacity arrives. Two ways it goes otherwise. If 14A converts one large Terafab participant into a named volume commitment, the external revenue line re-bases and 45 times 2027 looks cheap; if server supply loosens and October's list prices hold, the argument returns to $2.04 while the foundry keeps losing $2.1bn a quarter. The falsifier is narrow and dated: external foundry revenue still sitting near $293mn when 14A products ship in 2028.