Zero One Daily Intelligence Brief — June 9, 2026

Google's reported 3-million-chip order revives Intel's foundry as AI stocks rebound

Top Market Signals

Google order puts Intel's foundry back in play

Alphabet (GOOGL US) has reportedly placed an order with Intel (INTC US) for more than 3 million of its TPU (tensor processing unit) AI chips, with Intel manufacturing them from 2028. The report sent Intel up as much as 13% before the open.

According to The Information, cited by TechNews and Liberty Times, the order would be the first large external win for Intel Foundry, the contract-manufacturing arm Intel has struggled to fill since ceding process leadership to TSMC (2330 TT). A second data point landed alongside it: NVIDIA (NVDA US) is evaluating Intel's 18A (1.8nm-class) process for a part that fuses four GPUs into a single unit, though it has not yet committed. Tesla flagged in April that it plans to use Intel's next-generation 14A node for its Austin AI-chip site.

The signal cuts against the bearish read on Intel Foundry. Our 4 June Blade flagged Intel's CFO conceding that 18A had been overreach and leaning on 14A instead. A 3-million-unit Google order, if it firms up, is the external validation the foundry business has lacked, and it reframes Intel from a turnaround hope into a credible second source for hyperscaler silicon.

The caution is valuation, not demand. Intel Foundry is still lossmaking, yet the stock has run 464% over the past year and trades near 120 times 2027 expected earnings. Today's move prices in order wins that remain reports rather than signed contracts.

Linked stocks: INTC US, NVDA US, GOOGL US, 2330 TT

Sources: TechNews, 英特爾發了！傳Google 300萬顆TPU訂單

OpenAI files to go public after Anthropic

OpenAI has confidentially filed paperwork with the US Securities and Exchange Commission for an initial public offering, the company confirmed, joining its closest rivals in tapping public markets to fund the AI buildout. The filing came one week after Anthropic took the same step.

The move was reported across more than seven outlets including Bloomberg, AP and Reuters. OpenAI said the timing of any listing is not yet set. The filing lands in a crowded queue: SpaceX and Anthropic are both moving toward public markets, making the second half of 2026 a live test of whether listed investors will fund the AI capital cycle directly rather than through the chip and infrastructure names that have carried it so far.

For the listed supply chain, an OpenAI IPO matters less as a single stock than as a demand signal. OpenAI is a top customer for NVIDIA compute, a major Microsoft (MSFT US) commercial partner, and an anchor tenant for the datacenter and memory buildout. A successful raise would extend the funding runway behind the orders that Taiwan and Korean suppliers are already booking.

Linked stocks: MSFT US, NVDA US

Sources: OpenAI Files Confidentially for IPO, OpenAI joins Anthropic in IPO push

MediaTek limit-up leads Taiwan's AI-chip rebound

MediaTek (2454 TT) surged to its 10% daily limit at NT$4,475, rising more than 9% intraday and contributing over 200 points to the Taiex as Taiwan stocks bounced back. The chip designer, not index heavyweight TSMC, led the charge.

The trigger was its AI ASIC (custom application-specific chip) business. According to Liberty Times, MediaTek is reportedly winning a share of the lower and mid-tier custom-chip work outsourced by Broadcom (AVGO US), helping Google on its silicon, and progressing toward work with another large AI customer. Custom AI accelerators are the fastest-growing slice of advanced-chip demand, and a designer with proven smartphone-scale volume and TSMC packaging access is well placed to take share.

The move tracked a broad recovery. The Philadelphia Semiconductor Index had rebounded 5.61% the prior session, and Taiwan's market opened higher with financials adding support. TSMC, by contrast, stayed roughly flat near NT$2,295, leaving MediaTek and the mid-cap supply chain to drive the index.

Linked stocks: 2454 TT, AVGO US, GOOGL US

Sources: 聯發科訂單看俏 領軍台股反彈, 費半飆5.6%吹響反攻號角

Largan moves into AI optical interconnect

Largan Precision (3008 TT), Apple's main camera-lens supplier, told its annual meeting it will complete its first pilot production line for silicon-photonics CPO (co-packaged optics) components by September and fit it with automated equipment. The stock hit its limit at NT$3,885, up 9.90%.

CPO is the optical interconnect that AI systems increasingly need to move data between chips at high speed with less power. Largan's effort centers on the FAU (fiber array unit), the precision part that aligns optical fibers to a silicon photonics chip, where it said it can already hold tolerances below 0.3 microns. CEO Lin En-ping framed the line as validation rather than volume: once built, the company enters a verification phase, and moving from pilot to small-scale production would take a further 6 to 12 months.

This widens the optical supply base that recent Blades have tracked, from NVIDIA's TSMC-built CPO switches (3 June) to AUO's CPO push (6 June). An optics leader with Apple-grade precision moving into fiber alignment is a credibility step for the interconnect theme, even if revenue is years out. Largan's strength dragged optics peers Genius Electronic Optical (3406 TT) and several smaller lens makers to their limits.

The near-term business stays camera-led. April net income rose 88.5% to NT$1,831m, and May revenue grew 43% to NT$4,593m, with management guiding to full capacity in 3Q26 on the iPhone 18 ramp.

Linked stocks: 3008 TT, 3406 TT, NVDA US

Sources: 大立光光通訊FAU試產線在9月完成建置, 大立光股東會：產能利用率第3季拚滿載

Applied Digital signs 15-year US$5.2bn hyperscaler lease

Applied Digital (APLD US) signed a 15-year lease with a large US hyperscaler for 210 megawatts of compute capacity at its Delta Forge 2 campus, a contract it expects to generate US$5.2bn in revenue. The shares jumped 8.81% after hours to US$44.55.

The lease is structured take-or-pay, meaning the customer pays a guaranteed minimum regardless of usage. Applied Digital did not name the tenant but said it is the same investment-grade hyperscaler behind two earlier long-term leases, and that investment-grade cloud customers now account for about 70% of its contracted revenue. If all renewal options are exercised, the term extends to 30 years and cumulative revenue rises to as much as US$12.7bn. The site, in the southern US, uses zero-water cooling and is due online in 1Q28.

The structure is the signal. Fifteen-year take-or-pay commitments from investment-grade tenants show hyperscalers locking in AI compute capacity years ahead of need, which underwrites the power, cooling and datacenter buildout that Taiwan's server and component makers supply into.

Linked stocks: APLD US, NVDA US

Sources: TechNews

Cerebras jumps 18% as wafer-scale challenger re-rates

Cerebras Systems (CBRS US) climbed 18.32% to US$237.83 after its IPO quiet period ended and underwriters issued their first ratings, all bullish on the startup's wafer-scale approach to AI compute.

Cerebras runs its workloads on a single wafer-sized chip, its Wafer-Scale Engine, rather than stringing together many separate GPUs as NVIDIA does. The design cuts data-movement latency, which matters more as AI shifts toward heavier reasoning workloads. Its customers already include OpenAI and Amazon AWS. The ratings carry weight because of timing: underwriters are barred from publishing for 25 days after a listing, so the first reports after a quiet period often set the tone.

The stock remains volatile. Cerebras listed on Nasdaq on 14 May at US$185, spiked about 70% on day one, then fell more than 30% from its peak amid worries that AI names had run too far. Today's bounce reflects renewed conviction that low-latency, high-efficiency silicon is a genuine second axis of AI-chip demand alongside the GPU, not that the GPU leadership is under near-term threat.

Linked stocks: CBRS US, NVDA US

Sources: 華爾街看好 Cerebras 評等樂觀 股價勁揚18%

先聲 First Word — Exclusives from Chinese-Language Sources

ASE Suzhou packaging plant hit by worker strike

A packaging plant under ASE Technology (3711 TT) in Suzhou saw roughly 300 workers walk out on 6 June after pay negotiations broke down, with demands covering wage increases and reinstatement. A labor stoppage at an advanced-packaging site is an operational risk worth watching, given how tight back-end capacity is across the AI supply chain. (9 Jun 2026) Source: 日月光蘇州封裝廠勞工罷工約300人

Nanya deepens LPDDR tie-up with SK Hynix

Nanya Technology (2408 TT) said it will deepen cooperation with SK Hynix on memory as AI buildout accelerates, and expects memory supply to stay tight. The move extends Taiwan's push into the AI main-memory chain, where Nanya has been qualifying LPDDR for NVIDIA's server platforms. (9 Jun 2026) Source: 南亞科深化與SK海力士記憶體合作

Samsung's 2nm foundry orders set to jump in 3Q26

Samsung Electronics is reportedly pulling forward 2nm trial production to 3Q26, with additional foundry orders from two large customers potentially up around 130% year-on-year. A firmer 2nm order book would mark progress in Samsung's effort to close the leading-edge gap with TSMC. (9 Jun 2026) Source: 三星2奈米先進製程委製訂單3Q26將拉升