Intel's Packaging Paradox: A Rival's Copycat Move Can't Mask the Foundry Arithmetic

Intel's Q2 operational beat and TSMC's EMIB imitation boost shares, yet $11B GAAP loss and foundry losses reveal a fragile recovery.

When your fiercest competitor starts imitating your technology, it usually reads as vindication. For Intel, the news that TSMC is developing its own version of the EMIB packaging technology alongside partner Kinsus Interconnect Technology triggered exactly that reaction — shares surged roughly 13 percent in a single session, while TSMC itself climbed 7.6 percent. The market's logic was straightforward: if the dominant player in advanced chip manufacturing feels compelled to replicate Intel's silicon-bridge approach, then Intel Foundry's progress under CEO Lip-Bu Tan must be real.

The validation, however, arrived amid a stretch of extraordinary volatility that tells a more complicated story about investor confidence in the turnaround.

The Quarter That Confused Everyone

The packaging headlines landed just days after Intel delivered financial results that contained two sharply divergent narratives. On an operational basis, the second quarter was a blowout: revenue hit $16.13 billion, nearly 12 percent above analyst expectations, with earnings per share of $0.42 — double the consensus forecast. The client computing and physical AI segment accelerated from 1 percent growth in Q1 to 13 percent in Q2, reaching $8.88 billion. Data center and AI revenue surged 59 percent to $6.26 billion.

Yet under US GAAP, the company reported a staggering $11 billion loss, or $2.16 per share, driven by a $12.5 billion valuation write-down on trust shares tied to the CHIPS Act agreement with the US government. The market's response was telling: the stock fell nearly 8 percent the day after earnings, even though the operational numbers beat expectations across the board.

HSBC analyst Frank Lee captured the disconnect succinctly, noting that the market was reading this quarter entirely differently from the previous one. The gap between the GAAP loss and the operational improvement has become the central tension in Intel's story.

The Foundry Conundrum

The foundry business remains the fulcrum of the entire turnaround thesis — and its numbers are still sobering. Segment revenue grew 31 percent to $5.77 billion, but the operating loss deepened to $2.09 billion, translating to a negative 36 percent margin. More tellingly, only $293 million of that revenue — roughly 5 percent — came from external customers. The 18A production process is making meaningful progress, but Intel's contract manufacturing for third parties remains embryonic.

The broader operational picture is improving, however. The operating margin climbed to 11.1 percent in Q2, a dramatic swing from negative 24.7 percent in the year-ago quarter. On a trailing twelve-month basis, the margin reached approximately 7.6 percent, up from negative 8.3 percent. Inventory days fell from 137 to 118, signaling more efficient production planning.

A Balance Sheet Under Construction

Tan, now roughly 16 months into his tenure, describes his approach as a "bottleneck strategy" — concentrating the company on technological constraints like interconnect solutions, optical signal transmission, and voltage conversion, with all engineering teams reporting directly to him. The financial engineering has been equally aggressive.

Intel raised its full-year capital expenditure plan to $20 billion while absorbing a $14.2 billion hit to liquidity from the acquisition of its Irish joint venture, leaving cash at $29.73 billion. Debt rose 8.5 percent to $50.54 billion. The company has also leaned on strategic investors: Nvidia invested $5 billion in September 2025 — a stake now worth roughly $25 billion at current prices — while SoftBank has also taken a position.

The Packaging Race Intensifies

The EMIB technology at the center of the TSMC copycat reports uses silicon bridges to connect multiple chip components, enabling larger, more complex processors with greater efficiency and lower manufacturing costs. It has already attracted marquee customers: Nvidia, Google, and OpenAI are all using it, with Nvidia evaluating the technology for a future processor.

The competitive dynamics are sharpening. TSMC's own CoWoS packaging capacity is fully booked through 2026, with lead times stretching to 78 weeks. Intel, meanwhile, is pushing its EMIB-T packaging technology — which the company claims is roughly 50 percent cheaper than CoWoS — toward mass production starting in 2027. Yield on EMIB-T is approaching 90 percent, though substrate yield remains around 50 percent.

The reported order pipeline adds another layer of intrigue. Google is said to have ordered more than three million AI chips (TPUs) from Intel Foundry for 2028, while Nvidia may test Intel's 18A process for a multi-chip processor destined for its Feynman platform that same year. Tesla reportedly wants to use Intel's 14A process for its "Terafab" initiative. Intel declined to comment on any of these reports.

A Stock Caught Between Momentum and Caution

The market's ambivalence is visible in the charts. Intel closed the week at €78.42, down 0.78 percent on Friday and 3.45 percent for the week. The monthly decline approaches 30 percent, and the stock sits roughly 37 percent below its 52-week high of $124.58 from late June. It trades about 20.5 percent below its 50-day moving average, with a relative strength index of 39 — suggesting momentum has cooled despite the packaging news.

Bernstein's Stacy Rasgon has expressed his most optimistic view in some time, but the market's behavior tells a more cautious story. Investors appear to acknowledge the operational progress in foundry and AI while remaining wary of the segment's persistent losses. The company's Q3 guidance — revenue between $15.8 billion and $16.8 billion, with adjusted EPS of $0.38 — sits above consensus, but the market has shown it will punish the stock for headline losses regardless of the underlying operational story.

For all the excitement around TSMC's unexpected tribute, the fundamental question remains whether the reported hyperscaler interest in EMIB and 18A translates into concrete foundry contracts. Until external customers contribute meaningfully to the foundry segment's revenue, the gap between Intel's technological ambitions and its financial reality will continue to fuel the stock's dramatic swings. The company is also pushing forward with its Ohio manufacturing plans, paying overtime premiums to workers to bring 14A production online by 2031 — a reminder that the capital-intensive path to a full recovery still stretches years into the future.