Newsletter · · Ashutosh Agarwal

Foundry & Chip Equipment - Week of June 21, 2026: Intel's Apple Moment and a Memory Crunch With No Fix

Foundry and chip-equipment newsletter for the week of June 21, 2026. A Trump-announced Intel-Apple foundry tie-up sent INTC to a record even as CNBC anchors said they could not confirm it, while the durable story is a memory supercycle (prices up roughly 4x) that is supply-limited, not demand-limited, gated by clean rooms into 2027.

Foundry & Chip Equipment

Week of June 21, 2026: Intel's Apple Moment and a Memory Crunch With No Fix

For one strange Wednesday, Intel was the most exciting stock in semis. A presidential Truth Social post claiming Apple had agreed to design and build chips with Intel sent the stock to a record and dragged the whole foundry complex up with it, even as the two CNBC anchors closest to the story both said, on air, they couldn't confirm a thing. Underneath the noise, the more durable story is the one nobody can fix by year-end: a memory crunch so severe it's about to show up in the price of your next iPhone.

TL;DR

  • The Intel-Apple "deal" is a Trump post, not a signed contract, but the buy-side has expected Apple as an Intel foundry customer for a while, and the 18A-P production start gave the rally something real to stand on.
  • Memory is in "the strongest cycle in history." Prices up 4x in three quarters, an estimated $200/phone hit to Apple, and you can't pour a new fab in five minutes.
  • The whole chain is supply-limited, not demand-limited, ASML's CEO and Bernstein both frame 2026 as a "strong but constrained" year, gated by clean rooms, not orders.

What's new

1. The Intel-Apple announcement: loud, unconfirmed, and still a tape-mover. On CNBC's Squawk on the Street, 9AM hour (June 18), David Faber read Trump's post, "Apple's agreed to work with Intel to design and build its chips," and then said flatly, "I cannot confirm the Apple deal," with Jim Cramer adding, "I can't confirm it... there is no deal right now." Faber did confirm Intel's existing 18A foundry customers: AWS (in production), Microsoft, and Nvidia (pre-production). Treat the Apple headline as optionality, not fact, but the reaction was real: Intel rallied more than 10% on the day to a ~$675B market cap.

2. 18A-P actually entered production. On Schwab Network (June 17), Tom White flagged that Intel "began production of their most advanced chip node... the 18A-P," following 18A's move into PC chips in January. This is the substance beneath the Apple speculation, a better variant of a process that's finally yielding.

3. The memory cycle is the real macro story. On CNBC's Squawk on the Street, 11AM hour (June 18), reporter MacKenzie Sigalos laid out the math: memory prices up 4x in three quarters, a component-cost hit to Apple that CounterPoint pegs at $200 per phone, and Tech Insights estimating Apple may need ~$300 of price on a future Pro model to hold its near-50% gross margin. Cramer, on the 9AM hour, said he'd spoken to Tim Cook "at length", hyperscalers "swept everything," leaving Apple "no choice but to raise price."

4. ASML's CEO frames the whole thing as supply-limited. On Bloomberg Talks (June 17), Christophe Fouquet said the industry is heading into "a supply-limited market for AI, for semiconductor, for quite a few years," with customers pulling orders forward to "create longer-term visibility." He also noted that 80% of the world's advanced chips are bought by the United States, and pointed to Korean DRAM projects and a first Indian fab (Tata) coming online next year as incremental capacity.

5. The WFE up-cycle is gated by buildings, not demand. On TechSurge (June 16), Bernstein's Stacy Raskin, a former chip-equipment engineer, called 2026 "a pretty strong year for WFE... but a constrained year," because you need a clean room to put a tool in, and "they don't have the factory, so I have to build the factories first." Those clean rooms "start to come online next year." That's the read for AMAT/LRCX/KLAC: the ceiling this year is physical, and it lifts in 2027.

The debate

Bull case, TSMC's monopoly and a durable AI WFE up-cycle, was voiced loudly this week. Bloomberg Intelligence's Mandeep Singh (Bloomberg Tech, June 18) noted Nvidia has "already prepaid... $120 billion for TSMC capacity," and that even Apple, once TSMC's preferred customer, now feels Nvidia has the bigger lock-in. His read: "TSMC's monopoly is a problem for a lot of their big customers, including Apple," which is exactly why Apple is shopping. Sigalos reinforced it, TSMC's stock was up on the Intel news because it keeps a long backlog at its three fabs and earns fatter margins from hyperscalers than from Apple. Raskin and Fouquet round out the durability case: strongest memory cycle ever, AI demand still outrunning supply.

The Intel re-emergence sub-debate is genuinely two-sided. Intel CEO Lip-Bu Tan, in two long-form interviews (No Priors, June 18; The Angle from T. Rowe Price, June 17), made the operator's case: 18A in production, a 14A roadmap ("like 1.4 nanometer"), 10 and 7 already on the horizon, and a claim he hit a five-year turnaround plan "in 14 months." But he was also candid, "on the foundry side, we are very distant from TSMC... we have to be humble." The skeptics pushed back with numbers. Wedbush's Matt Bryson (Neutral, $60 PT) on the 11AM hour: "Taiwan Semi has well over 60% gross margins... Intel, we're still around 40," and "TSM is much better at Foundry than Intel is." Cramer was blunter, "the real foundry doesn't come to 2028... it's not a place to look to for profit yet." Raskin's verdict on TechSurge split the difference: the 18A process "if they can get it to yield, is good," but the yields "are still not good" on margins, and competitive server CPUs wait for Coral Lake in ~2028.

What wasn't voiced this week: the classic bear cases, Taiwan Strait/geopolitical concentration risk, a China pull-forward unwind, and mature-node digestion, simply weren't argued on the pods this week. Worth flagging given the up-cycle euphoria, but not something I'll attribute to a source that didn't say it.

Names in play

Intel (INTC) is the obvious one, up more than 250% year-to-date per Sara Eisen, now trading on EV/sales "at the same levels as both AMD and Taiwan Semi," per Bryson, "trying to beat both... but not being there yet." Next catalyst: whether the Apple relationship gets confirmed in the second half, where Tan has hinted at "a couple of big irons in the fire." Micron (MU) is the memory derivative, Cramer noted it's up "800-plus percent" over twelve months and reports earnings this coming week, with Bryson warning that pricing "ticked up again in an unexpected manner," so "everyone's numbers have to move up again." Apple (AAPL) sits on the wrong side of the memory squeeze and the wafer-supply game, losing TSMC preferential treatment to Nvidia, eating a $200/phone component hit, and reportedly leaning on Intel and Samsung as second sources.

Read-throughs

  • Advanced packaging / CoWoS: still the bottleneck. Raskin flagged how much CoWoS capacity customers are reserving; Tan is pitching Intel's "EMT" as a next-gen alternative and spreading bets across new materials (silicon carbide, indium phosphide), glass substrates, and even artificial diamond. HBM is the pull, roughly 85% of an AI chip's silicon area, and it burns ~4x the silicon of standard DRAM at lower yields (Raskin, TechSurge).
  • EUV / litho: Fouquet's "supply-limited for years" framing and confirmed order pull-forward is the cleanest positive for ASML and the deposition/etch/metrology names behind it, though he gave no High-NA booking counts or 2026 guide.
  • Fabless wafer-price exposure: Apple is the marginal loser; Nvidia's $120B prepayment looks like the smart pre-buy.
  • China toolmakers: only a generic mention this week, Raskin noted Chinese WFE players are "pretty good... not as good as the U.S. guys," gaining share mostly because export controls box out the incumbents. No SMEE/Naura/AMEC/Piotech specifics surfaced.
  • Quantum (frontier read): On Lead-Lag Live (June 15), WisdomTree's Christopher Gannatti detailed a ~$2B US government commitment across nine quantum names, with IBM taking $1B and matching it to build a superconducting-qubit fab ("Andurran"), targeting error-corrected quantum by 2029. The hardware read-through is a cryogenic supply chain, niobium, helium isotopes, dilution refrigerators, that rhymes with semi manufacturing. Pure-plays cited: D-Wave, IonQ, Rigetti; Quantinuum just IPO'd. Long-dated, but the picks-and-shovels framing is the one to watch.

What changed

The center of gravity flipped to Intel. After quarters of Intel-as-value-trap, this week's tape had two CEOs (Tan, Fouquet) and the most credible sell-side voices (Raskin, Bryson, Singh) all engaging seriously with an Intel-re-emergence scenario, even the skeptics now argue valuation and timeline, not viability. That's a different debate than we were having a quarter ago.