Here’s how Liwei’s publicly disclosed US and Taiwan stock buys have performed since the August comeback on X. The timing on the US buys has been much better than on the Taiwan buys. Liwei doesn’t announce buys often. But when Liwei says “I buy or I bought” it’s usually because Liwei thinks the stock is genuinely cheap. You can use Grok to verify all of this against Liwei’s public posts. $INTC $FORM $AEHR $MRVL

According to Jensen: NVIDIA invests about 20% in design and 80% in verification; currently, most AI labs do the exact opposite, investing 80% in design and only 20% in safety, verification, and evaluation. if we estimate based on computational resources invested, while keeping the design investment unchanged, simply bringing safety, verification, and evaluation up to the same proportion would require total compute to increase to 4 times the original amount. 🤯🤯🤯

Two of the 'Big Four' on the table are Taiwanese! ❤️

ECOC 2026 – Vertical Integration Is $COHR ’s Key Advantage： According to Citi, Coherent sees up to $15K in optical content around a 100T chip by supplying the complete PhotonLink solution (assuming 100 Tbps I/O and 200 Gbps optical lanes). This turnkey package spans the external laser source, laser-to-chip interconnects, chip-to-faceplate assemblies, and packaging/testing. Management expects scale-out CPO to ramp in Q4 2026, with scale-up CPO and NPO following in H2 2027. $LITE $COHR $AAOI

ANTHROPIC SIGNS $11.6B AKAMAI CLOUD DEAL Anthropic has committed $11.6B over seven years to use Akamai $AKAM infrastructure for growing CPU workloads, with options to expand the relationship by another $9B, bringing the potential total to ~$20B. As part of the deal, Anthropic received warrants that could give it up to ~5% of Akamai. About 2% is expected to vest with the initial commitment, with the remaining 3% tied to further cloud spending. Akamai expects ~$5.5B of capex related to the initial deal, including an additional ~$1.7B in 2026 to secure components such as memory.

testing testing testing @RYANHINGSHING @photonjaeger @LIWEI_TWCapital @cherryPayment

The CPU boom is giving testers more work. In July, Advantest raised its 2026 SoC tester market forecast midpoint from $9.1B to $11B. UBS’s Sept. 22 tech-tour note adds a longer-term driver: Advantest believes CPU test cycles could approach GPU levels over the medium to long term as advanced packaging adoption grows. $FORM $TER

“Packaging is really important, by the way, because if you look at it, the packaging capacity is like non-existent.”— Elon Musk, speaking with Gwynne Shotwell on All-In. (podcasts.happyscribe.com) Elon ignited the advanced packaging narrative again, and Liwei's pick ASE ($ASX) is knocking on the door of an All-Time High. (Liwei’s picks rarely disappoint 😉) I want to highlight ASE ($ASX) again for anyone who missed the initial run. Think of it as a mini-$TSM, and one of those rare names that international investors can buy directly through its ADR. The chart is printing a textbook cup-and-handle. If we get a quick handle pullback before the breakout, that’s your prime entry window—because once it clears ATH, there’s no looking back. Here are 15 concrete catalysts behind the ASE($ASX) bull case: 1. $TSM says packaging shortages are limiting customers’ growth. That supports the opportunity for ASE to absorb additional outsourced work. (investor.tsmc.com) 2.TSMC and ASE co-chair Taiwan’s 3DIC manufacturing alliance, coordinating packaging processes, equipment and materials. The partnership already has an operating framework. (money.udn.com) 3. $AMD officially named ASE and SPIL as partners developing and qualifying wafer-based 2.5D EFB packaging for Venice server CPUs. (ir.amd.com) 4.Taiwan’s Economic Daily News reported additional $GOOG's TPU packaging and testing orders for ASE in August. A customer catalyst beyond GPUs, pending company confirmation. (money.udn.com) 5.The same report described expanding Intel outsourcing tied to EMIB. ASE could benefit across competing packaging platforms. (money.udn.com) 6.ASE is expanding full-process advanced packaging. Taking on more manufacturing steps creates an opportunity to capture more revenue per package. 7. Management expects 2026 leading-edge packaging and testing revenue above $3.5 billion, with a target to double it in 2027. (cna.com.tw) 8.Planned 2026 capex increased from $8.5 billion to roughly $10.5 billion, adding both facilities and equipment to support demand. (cna.com.tw) 9. Seven factory buildings are under construction simultaneously in Kaohsiung. Capacity expansion is already underway. (money.udn.com) 10. ASE’s automated 310×310 mm panel-packaging platform is expected to begin production by Q1 2027, supporting FOCoS and FOCoS-Bridge. (ase.aseglobal.com) 11. Silicon-photonics/CPO-related production is targeted for late 2026, initially at small volumes. Another potential revenue stream is approaching commercialization. (udn.com, fool.com) 12.The Renwu testing project with WinWay and HTT targets phased operations in 2027, expanding wafer and chip testing capacity. (ase.aseglobal.com) 13.August packaging, testing and materials revenue reached NT$51.29 billion: +53.1% YoY in Taiwan dollars and +41.4% in US dollars. (prnewswire.com) 14.Q2 packaging, testing and materials revenue grew 36.3% YoY. Segment gross margin rose to 27.3%, from 26.0% in Q1. (sec.gov) 15. Management guided Q3 segment revenue up another 11–13% sequentially in Taiwan dollars, with gross margin improving to 28–29%. (cna.com.tw) The $ASX bull case: capture more packaging and testing work per AI system, then convert new capacity into profitable revenue. The company that finishes the chip when TSMC’s advanced-packaging lines are full, when $AMD needs EFB, when $GOOGL / $INTC need a second packaging path, and when test times keep getting longer. That’s why the multiple re-rated. That’s also why a pullback into strength is more interesting than chasing the first tick through $45.52. Also, the legendary @PhotonCap has been pounding the table on advanced packaging lately. Make sure to check out and subscribe to his Substack: substack.com/@photoncap/note…

Imagine if one billion people were running multiple personal AI agents from different brands like $META $GOOG $AAPL, that’s a lot of CPUs and memory...

CPO bottleneck: Test, test and still test…. @photonjaeger @JoeyOnMarkets Everyone will reach a consensus like us sooner or later 🫡 $FORM $TER $KEYS $AEHR $VIAV

AI Infra’s Next Trade: Taiwan–US Air Cargo? 🤯🤯 Central News Agency reports that $2610.TW’s Americas cargo chief estimates its 2026 cargo revenue will grow 40–50%. North America already accounts for more than 60% of its cargo revenue. $2618.TW says bulky AI equipment is driving requests for charter flights and reserved cargo space. AI servers need to reach data centers, and some loads need dedicated freighters. Could airlines cash in on the buildout too? This is starting to feel like 2021 🤯 Sources money.udn.com/money/story/56…

Phison (群聯）8299.TW CEO Warns Memory Shortages Could Outlast 2029 So who is the clown 🤡? Jason Chen or K.S. Pua? Commercial Times reports 8299.TW CEO K.S. Pua sees DRAM and NAND supply remaining insufficient even as Chinese producers add capacity. His reasoning: if AI demand keeps growing, memory expansion will struggle to match the computing chips that need it. For device makers, that could mean prolonged component-cost pressure. Pua warns that rising memory prices could weaken PC and smartphone sales over the next one to two years, delaying replacement purchases. This is his outlook, not a confirmed shortage timeline. Sources ctee.com.tw/news/20260924700…

H/T: facebook.com/share/1EyxQGVR8…

$SKHY broadens its venture focus to optical links and software 👀 DIGITIMES reports that the Korean memory maker launched a corporate venture-capital brand and expanded its investment scope to include optical interconnects, system software, AI computing and data centers. For SK Hynix, that creates a route to technologies that connect and manage the systems surrounding its memory products. The strategic implication is broader participation in AI infrastructure Sources digitimes.com/news/a20260923…

TPUs still make up the bulk of compute today, and supply chain bottlenecks remain pervasive. However, the rise of $Meta’s Muse could begin to threaten $GOOG Search's position as the primary web gateway. With a complete ecosystem spanning Search, Chrome, Android, Gmail, and YouTube, Google won't take this lying down. If Google accelerates its agentic push, it won't just be about TPU inference—underlying orchestration, APIs, browser automation, and VMs will significantly drive up CPU workloads as well. That’s why I’ve recently started scouting Google’s CPU supply chain. It’s still an early angle, but definitely worth digging into.