Delta Electronics, the power electronics giant, will spend NT$70 billion (US$2.2B) on capex this year, up from NT$46 billion last year and sees the 2nd half of the year even better than the first half. Mass production of 800V DC equipment is expected to begin in the 3rd quarter, said Chairman Ping Cheng, also: “The wave of AI data center construction will not stop.” #DeltaElectronics# #DC#

Weekly|AI Slowdown by Sector, GPT-6 Astra Computer Use, Delta VPD Re-accel, Enterprise AI Vol.3, SAIL FY27Q2, EMC Price Hikes The tape last week split again: broader indices leaned soft into PPI/CPI and the next FOMC, while semis held up on company-specific news, where SOX finished higher even as mega-cap tech traded mixed. That is the same relative setup we have been tracking since memory broke out: hardware names responding to incremental AI infrastructure demand while the index still waits on the macro prints. The research week was dominated by compute demand from a different angle. GPT-6 Astra is OpenAI’s first flagship with computer use as a core capability, and we read it as the Claude 3.7 moment for knowledge work: the demos moved from terminals into CAD, Excel, Blender and tax forms, and the addressable pool is an order of magnitude larger than coding. Multi-step reliability is still the gap, but the trajectory is visible, and OpenAI’s RSI note put numbers on the loop already turning inside the labs: agent hours at 3.1x human hours, with compute as one of the constraints that tightens as other bottlenecks ease. That framing matters more for the next pre-training cycle than any single benchmark. Separately, the Anthropic-led slowdown debate is back on X. Our take is that release cadence is not the same as training cadence, and a single brake would miss how uneven the stack already is. Coding, math and cyber are remaking workflows where RL data and feedback loops exist; most other industries still lack workflow traces and a usable context layer. Finance is the familiar case. GPT chats and Notion notes are easy to log, while the research-to-EPS judgment path is not. If policy slows anything, the useful version is sector by sector, with more effort lifting the laggards than cutting the frontier to match. Power and enterprise spend filled in the rest of the stack. We initiated on Delta: 2H26 re-acceleration is AI PSU plus VPD, with DC/DC still underweighted by the market at an estimated 8.3%/12.2% of revenue in 2026/2027 and >70% VPD share at $400-600 content per TPU. Enterprise AI Vol.3 went out with ten more samples: spend is still growing, but the questions have shifted to who needs the expensive models, whether time saved converts into revenue or real cost cuts, and how budgets get reallocated once caching, defaults and local models bite. This Week’s Reports AI Slowdown by Sector — release pace is not training pace, and a uniform brake misses how uneven the stack already is. Our debut Column argues coding, math and cyber remake workflows where RL data and feedback loops exist, while most other industries still lack workflow traces and a usable context layer; if policy slows anything, the useful version is sector by sector, lifting the laggards rather than cutting the frontier to match. GPT-6 Astra — computer use is the Claude 3.7 moment for knowledge work, and RSI is already turning the compute loop inside the labs. Astra is the first flagship with computer use as a core capability, OSWorld at 72.6%, and a visible path from coding’s trillion-dollar pool toward white-collar work an order of magnitude larger; OpenAI’s RSI note put agent hours at 3.1x human hours with compute as a binding constraint. Delta Electronics — 2H26 re-acceleration is AI PSU plus VPD, and DC/DC is still underweighted. We initiate on as a grid-to-core AI power name: VPD takes DC/DC to about 8.3%/12.2% of revenue in 2026/2027 with >70% share and $400-600 content per TPU, while we model FY27/FY28 revenue growth of 53%/28%. Enterprise AI Vol.3 — spend is still growing, but the next budget round is gated by usage depth, model tiering and measurable ROI. Ten more enterprise samples show broad basic access with highly concentrated heavy users, production agents in a few names, and cost controls (defaults, caching, local offload) that often get reinvested rather than cutting the total AI envelope. Premium Report Snapshot A portion of our research is reserved for Premium subscribers and is not distributed via Substack. Below is a snapshot of what Premium subscribers received this week beyond the Substack feed. Deep | EMC: Price Hikes Continue into 26Q4; High-End Leadership Intact Preview & Review | $SAIL FY27Q2: Headline Results Broadly in Line; AI Commercialization Better Than Expected Weekly Expert Interviews Summary A snapshot of the expert interviews we conducted during the past week is below; full transcripts and takeaways are available on the FUNDA platform. Google — TPU v8 Internal Priority and v7 Price Pressure (GOOGL, AVGO) Astera Labs — Scale-Up Interconnect and Memory Expansion Roadmap (ALAB, AMZN, MRVL, CRDO) Bloom Energy — Manufacturing Capacity Expansion and Service Risk (BE, NBIS) SailPoint — AI Identity Products Broaden Growth Sources (SAIL, OKTA, MSFT) Enterprise AI — API Migration Reshaping Cost and Workflows PCB/CCL Supply Chain — PTFE Backplanes and Material Bottlenecks (NVDA) Legal AI — Spend Growth, Productivity ROI and Headcount Pressure AI Servers — Rack Power Architecture and Cooling Mix (AMD, ORCL, META) Detailed Report

Deep|Delta Electronics: 2H26 Re-acceleration on Deck, VPD Still Overlooked Delta Electronics ( the broadest play on AI power spanning the grid-to-core value chain. We expect 2H26 revenue to re-accelerate on stronger AI PSU demand and explosive VPD module growth, with GPM expanding from 26Q4 on a better mix, economies of scale, and price pass-through, and OPM improving even faster. The market, in our view, has overlooked Delta’s DC/DC business: we estimate it jumps to ~8.3%/~12.2% of total revenue in 2026/2027 on the VPD ramp, where Delta holds >70% share at US$400–600 of content per TPU. At a ~50% GPM, the mix shift should more than offset HVDC’s margin dilution. Liquid cooling should grow another ~80% in 2027, led by AWS’s IRHx, while HVDC enters volume production in 26Q4; fears of a timeline delay are overdone. With SST and SOFC optionality layering in from 2028, we model revenue growth of 53%/28% in FY27/FY28 and EPS of TWD77.3/102.4. The stock trades at ~17.8x FY28E EPS (based on the Sep 4 close of NT$1,825). Detailed Report

$NVDA and $GOOGL lead 800V DC ahead of schedule. "Ahead of schedule", pulled up to Q3 2026 with small volume shipments starting . - Delta Electronics (2308), $VRT - Song Chuan Precision (7788) - Schneider Electric, Eaton, Siemens. All flagged as beneficiaries. "Market sources indicate that Nvidia’s Vera Rubin platform and Google’s next-generation AI data centers will be the first to adopt the technology" Source: Commercial Times The power semi trade should be happy to hear this.

AI rack battery backup units are moving from 3.5kW to as much as 25kW in two years. Taiwan battery module suppliers say 3.5kW was the mainstream BBU output in 2025. That moved to 5.5kW in 2026. Dynapack (3211 TT) has started sampling units above 15kW and targets mass production in 2027. Celxpert (4931 TT) plans 15kW and 25kW models in 2027. Dynapack also plans to triple its BBU capacity this year. $NVDA’s MGX-compatible 800 VDC power rack arrives in 2H26. A Delta Electronics (2308 TT) design shown at GTC pairs a 110kW power rack with five BBUs. The suppliers say higher-output modules require more battery cells, extending backup time and raising the selling price per BBU.

Two VistaShares Supercycle ETFs. One powers the AI revolution. The other powers the grid that runs it. $AIS vs $POW VistaShares Artificial Intelligence Supercycle ETF $AIS: Launched: December 3, 2024 Assets Under Management: $855.3M Expense Ratio: 0.75% 30-Day SEC Yield: -0.26% 1-Year Return: +215.95% 🟢 Top Holdings: 🥇 SK Hynix 12.05% weighting 🥈 Micron Technology $MU 8.11% 🥉 Advanced Micro Devices $AMD 4.41% VistaShares Electrification Supercycle ETF $POW: Launched: October 28, 2025 Assets Under Management: $67.3M Expense Ratio: 0.75% 30-Day SEC Yield: 0.26% 1-Year Return: N/A — launched October 2025 Top Holdings: 🥇 LS Electric 8.84% weighting 🥈 Delta Electronics 7.24% 🥉 Powell Industries $POWL 6.48% Performance so far in 2026: $AIS: +102.97% 🟢 $POW: +58.10% 🟢 Disclosure: @VistaShares is a WOLF Financial partner. This is for informational purposes only, not financial advice.

Weekly|SNOW Data Flywheel & CoCo, Optics Scale-In, BE Delivery Debate, Enterprise AI Vol.2, CRDO/MDB/PANW The software vs semi split that defined last week finally started to rebalance. Software that printed what the channel had been pointing to kept working through most of the week, while the semi tape spent the first half getting tagged by rates — the 10-year pushed to 4.8% on Tuesday, the highest since January 2025, after Warsh’s hawkish Jackson Hole turn had traders pricing hikes rather than cuts. By Friday the setup flipped: semis clearly outperformed, and memory names broke out. Near-term software short cover looks largely done, the outperformance looks stretched, and the momentum unwind in semis has already gone to an extreme — which leaves less of that particular overhang in the names that got hit hardest. On the software prints, SNOW was the cleanest expression of the channel read. We went into the print cautious on 2Q because of tough comps and more constructive on 3Q, and the flywheel was stronger than even that. Product revenue grew 37% year over year, a 5.2% beat that matched last quarter, and the full-year product-revenue guide jumped from 31% to 36%, one of the largest annual raises in the company’s history. CoCo is no longer a product call. Accounts passed 9,100 with more than 2,000 net adds in the quarter, CoWork is at 5,800, and customers are consolidating transactional data onto the platform. Our preview had CoCo adding about two points this quarter; the print confirmed the direction and then some. The data flywheel is what we think the multiple should be underwriting, not a single SKU. Optics is the other narrative that actually moved, and it moved inward. Scale-in, as Lumentum’s CEO put it, is the thing to watch over the next 12 to 18 months: optical links inside the tray, targeting memory-class bandwidth, roughly 10x scale-up per GPU. NVIDIA has already written $2bn checks into Lumentum and Coherent; Marvell’s Celestial AI deal and the Google warrant, which discloses memory interface controllers and near-memory compute, are the capital trail. CRDO’s print sits in the same stack from the other end: a record quarter, Q2 guide above the Street, optics still a 2H ramp. On power, the AIDC demand debate is over. Bloom’s multiple has already reset from about 85x to 42x forward EV/EBITDA after the July shorts, and the stock has not re-rated on the 2Q beat-and-raise. We think the question is now whether Bloom can convert booked slots into shippable MW. We model about 2.7GW of year-end 2027 manufacturing capacity and 2.2GW of 2027 deliveries against 2.0GW consensus. Scandium looks manageable; service capacity is the overlooked constraint. Enterprise AI is splitting the same way the tape is: Vol.2 went out on Substack this week, spend is still growing, and the samples have diverged. Production-tied use cases keep adding; per-head allowances are getting cut. The increment is in APIs and production workflows, and ROI only converts into a budget when hours saved cross a threshold large enough to reallocate a person, about 30%, which is why some CFOs are already holding the total IT envelope flat. This Week’s Reports Optics — scale-in puts optical links inside the tray, and the bandwidth it targets is roughly 10x scale-up per GPU. Lumentum’s CEO called it the thing to watch over the next 12 to 18 months; NVIDIA’s $2bn checks into Lumentum and Coherent, and Marvell’s Google warrant covering memory interface controllers, are the capital already moving. Preview & Review | SNOW FY27Q2: Beyond a Single Product, a Data Flywheel That Merits a Longer Horizon. Product revenue grew 37% YoY and beat sell-side by 5.2%, with the full-year product-revenue guide raised from 31% to 36%. Our preview had CoCo still very strong and called 35% growth, a 4-point beat versus guidance; the print confirmed the CoCo-to-flywheel read, with CoCo accounts above 9,100 and CoWork at 5,800. Bloom — the AIDC demand debate is over; delivery execution is the rerating variable. Production slots look booked through 2028, we model about 2.7GW of YE27 manufacturing capacity and 2.2GW of 2027 deliveries against 2.0GW consensus, and forward EV/EBITDA has already reset from about 85x to 42x after the July shorts. Enterprise AI Vol.2 — spend is still growing, but the increment is in production workflows and ROI is gated by hours saved. A second set of seven enterprises independently corroborated Vol.1: seats are maturing, APIs and production workflows take the increment, and scattered 5-10% time savings do not cut cost until effort on a class of work falls by about 30%. Premium Report Snapshot Below is a snapshot of what Premium subscribers received this week beyond the Substack feed. Review | CRDO FY27Q1: Record Quarterly Revenue, Q2 Guide Above Street; Optics Ramp 2H Review | PANW FY26Q4: Organic Growth Modest Re-Accel, but vs Raised Buy-Side Bar Not a Clean Beat Deep | Delta Electronics ( 2H26 Re-Acceleration; VPD Overlooked Weekly Expert Interviews Summary A snapshot of the expert interviews we conducted during the past week is below; full transcripts and takeaways are available on the FUNDA platform. MongoDB — AI Production Scaling and Agentic Usage Growth (MDB, SNOW, GOOGL) Marvell — DSP Competitiveness and Optical-Engine Scale Bottlenecks (MRVL, AVGO, NVDA, SMTC) NVIDIA — Co-Packaged Optics Ramp and Supply Bottlenecks (NVDA, LITE, TSM) AWS — Track-Optimized Networking and Active Cables (AMZN, CRDO, NVDA) AI Data Centers — Gas Turbine Supply Constraints and Service Economics Palo Alto — Platform Consolidation and Growth Acceleration (PANW, CYBR) Zurich Insurance — Enterprise AI Spend and Adoption Economics (MSFT, GOOGL) Top Logistic Firm — Governance-Driven AI Spend Expansion (MSFT) Detailed Report