# Research run 2026-09-27T23:43 UTC ## What changed - Added Tesla, BYD and Toyota to the watchlist at the owner's request. - BYD accumulate: August NEV sales +18% YoY, overseas +134.5%, domestic China -14.3%. - Tesla watch: Q3 delivery print 2 Oct, Goldman cut to 435k against 454k consensus. - Toyota watch: 8.6x earnings, near the bottom of its 52-week range. ## Market summary The S&P 500 closed Monday at 7,620 (-0.5%), about 1.4% below Aug 27, and the Fed is near-certain to hike at 4am AEST Thursday: futures price roughly 90-93% odds of 25bp to 3.75-4.00%, the first hike since 2023, versus 34% priced on Aug 27. The repricing came in three steps: Warsh used Jackson Hole (Aug 28) to demand disinflation 'clearly and at sufficient speed', August payrolls (Sep 4) beat at +162k with unemployment steady at 4.1%, and August CPI (Sep 11) printed 0.4% m/m and 3.4% y/y with core 0.3% m/m, above forecasts. The 10-year touched 5.01% Monday, highest since October 2023 (4.68% on Aug 27), the ECB hiked 25bp to 2.50% on Sep 10, and Brent is near $107 with WTI $102.7 after Houthi attacks on Saudi Arabia and a pipeline strike Riyadh says could disrupt up to 4% of global supply. Chips took the extra hit Monday: SOX -5.9% and NVDA -3.4% to $210.96 after Amodei, Altman and Musk called for slowing AI development, though the S&P ended well off session lows and NVDA ticked up after hours. Asia is mixed at Tuesday lunchtime: Nikkei +0.9%, Kospi +0.3%, Taiex -0.1%, Hang Seng -0.3%, ASX 200 -0.9% at 8,669; China's August industrial output beat at +5.2% y/y while retail sales missed at +0.4%. Ahead in AEST: US retail sales 10:30pm Wednesday, FOMC decision 4am Thursday with Warsh's press conference ~4:30am, BOJ Friday (25bp to 1.25% expected), then the RBA on Sep 29 with a hike about 74% priced (cash rate 4.35%; NAB calls September, CBA and Westpac November). ## Watchlist changes - added TSLA:US (watch, medium conviction) - added 1211:HK (accumulate, medium conviction) - added 7203:JP (watch, medium conviction) ## Watchlist - ALAB:US Astera Labs - accumulate (medium) - Astera Labs sells the connectivity silicon (PCIe/CXL retimers, Scorpio fabric switches, Ethernet) that ties GPUs and XPUs together inside AI racks; Q2 2026 revenue was $392M (+104% YoY, +27% QoQ) with eight straight EPS beats, and Scorpio switches become the largest product line in Q3, a quarter early. The 18-day gap was rough on the tape, not the story: shares spiked ~11% on Sep 4 on S&P 500 inclusion bets, missed out when Bloom Energy, Illumina and Everpure were picked instead, presented at Citi's Global TMT conference on Sep 9 reiterating Scorpio X as the major revenue driver, then fell 11.7% on Sep 14 to $257.04 in the market-wide AI-slowdown selloff after Amodei and Altman called for pacing frontier development. Now -15% vs Aug 27 and ~48% below the $498 YTD high with no company-specific negative in the gap; Q3 (Scorpio becoming the biggest line) is the proof point. - AVGO:US Broadcom - accumulate (medium) - Broadcom is the leading custom AI accelerator (XPU) and AI networking supplier to hyperscalers, and the Sep 2 print validated the franchise: Q3 FY26 revenue $29.6B (+86% YoY), AI semiconductor revenue $16.7B (+221% YoY, +54% QoQ), non-GAAP EPS $3.32 (+96%), with a Q4 guide of ~$34.8B total revenue (+93%) at ~66% non-GAAP operating margin. The market sold the guide anyway - the ~$21.7B Q4 AI revenue figure was read as light against buyside hopes and shares fell over 6% after hours - and the stock then lost another 4.8% to $344.72 in the Sep 14 AI-slowdown selloff, leaving it -6% vs Aug 27 and ~28% below the June high. Hock Tan went on CNBC on Sep 14 to say AI revenue targets have not changed, and the VMware Explore announcements (Private AI Cloud, AI Factory, AgentMinder) keep the software annuity compounding. Path to >$100B of AI revenue in 2027 restated; the bar, not the business, is the problem. - CRDO:US Credo Technology - accumulate (medium) - Credo makes the active electrical cables (AECs), optical DSPs and retimers that link GPUs and switches inside AI clusters, and its ZeroFlap AECs are the default in-rack copper link for hyperscalers. The Sep 1 Q1 FY27 print was a record - revenue $479.0M (+114.7% YoY, +9.6% QoQ), GAAP net income $129.4M, non-GAAP EPS $1.20, non-GAAP gross margin 68.0% - and the Q2 guide of $525-535M implies over 85% YoY growth. But the margin guide (non-GAAP 67-69%, GAAP 62.9-64.9%, below Q1) crystallised the feared reaction: the stock fell ~20% on Sep 2 as analysts cut targets, kept sliding, and lost another 7.9% on Sep 14 in the AI-slowdown selloff to close at $150.09. That is -37% vs Aug 27, ~51% below the 52-week high and roughly 33% down over three months, while consensus sits near $281 and the demand base (four hyperscalers each over 10% of revenue) is unchanged. We read this as a valuation and margin-mix reset, not a demand break - the most oversold name on the list. - SKHY:US SK hynix - accumulate (medium) - SK hynix supplies more than half of the world's HBM (about 58% of HBM revenue share) and posted record Q2 2026 results: revenue KRW 79.3T (+257% YoY) and operating income KRW 60.5T (+557%) at a 76% operating margin, with HBM4 mass shipments ramping in H2 and a KRW 40T buyback running through November. It was the strongest name on the list through the gap: the ADR closed Sep 14 at $175.63, +7.7% vs Aug 27 even after falling 7.6% that day in the AI-slowdown selloff, because the memory shortage keeps tightening - DRAM inventories are reported below 10 days and BofA raised its target to $268 from $250 on Sep 14. The two overhangs are resolving: the company tabled a revised wage deal on Sep 11 lifting the cash share of performance bonuses from 40% to 50%, with a company-wide union vote Sep 15-16; and the Indiana packaging site is under way. Samsung and SK hynix also rejected Kepco's $19B power-prepayment proposal, a sign of bargaining power, and Solidigm pre-IPO reports drew a 'nothing determined' response. - 3750:HK CATL - buy (high) - CATL is the world's largest battery maker and the clear leader in grid storage: H1 2026 revenue rose 55% to RMB 277B and net profit 42% to RMB 43.3B, with energy storage revenue up 88% (nearly 20% of sales), and Jan-Jul 2026 global EV battery share was 39.9% vs BYD's 14.7%. The gap was all sentiment, no fundamentals: the shares ground down to a six-month low of HK$546.5 on Sep 11 'amid a storm of negative factors' and closed HK$549.00 on Sep 14 (~$70, -11% vs Aug 27) as BYD's weak H1 (profit -20%) soured China EV sentiment, Washington attacked Ford's CATL licensing tie-up, and Li Auto said it will shift its lineup to in-house batteries. Against that, Goldman initiated at Buy with a HK$946 target, Q2 profit rose on surging storage demand even with soft China EV volumes, and the stock now trades at ~19x earnings - the cheapest quality name on this list for the AI-era storage buildout. - BE:US Bloom Energy - accumulate (medium) - Bloom's solid-oxide fuel cells are a mainstream on-site power solution for AI data centers that cannot wait for grid interconnects: Q2 2026 revenue jumped 166% to $1.07B with a 33% gross margin, full-year guidance is $3.9-4.2B revenue, and the 2.8 GW Oracle agreement plus Nebius Vineland show hyperscalers and neoclouds in the customer base. The big gap event: S&P Dow Jones announced on Sep 4 that Bloom joins the S&P 500 effective Sep 21 (with Illumina and Everpure), triggering an ~11% jump and index-fund buying; Mizuho raised its target to $351 from $242 on stronger pricing and demand. The stock closed Sep 14 at $257.05 - down 6.8% in the AI-slowdown selloff but still +16% vs Aug 27 and the best performer on the list over the gap. Next earnings moved up to Oct 27. - ENR:DE Siemens Energy - accumulate (medium) - Siemens Energy is Europe's answer to GE Vernova: gas turbines, grid technologies and wind, with record fiscal Q3 orders (EUR 17.9B, book-to-bill 1.57), a EUR 162B backlog and Grid Technologies' backlog at EUR 51B on US data-center transformer demand. The Transformation of Industry separation is progressing: the unit is being readied under a new standalone identity reported as 'Omterra', private equity is circling the EUR 5.7B-revenue business (Goldman running the process; earlier reports pegged the valuation above EUR 10B), and the sell side remains lined up behind the plan with consensus near EUR 196. The stock had held up until Monday: Xetra close EUR 132.66 on Sep 14, -8.0% on the day as European markets slid on the AI-safety warnings and Fed rate-hike bets, leaving it ~$153 in USD terms, -13% vs Aug 27. Nothing in the gap changed the sum-of-parts case; the AI-power complex simply de-rated together. - GEV:US GE Vernova - accumulate (high) - GE Vernova is the key bottleneck supplier of gas turbines and grid equipment for the AI power buildout: Q2 2026 orders rose 88% organically to $24.2B, total backlog ~$176B, gas turbine backlog 116 GW (targeting 125 GW under contract by year-end), and H1 free cash flow of $5.1B already exceeded all of 2025. Gap news was constructive: on Sep 3 it committed with Studsvik, Samsung C&T and DS Investment Partners to advance a four-unit, ~1.2 GW BWRX-300 small-modular-reactor project in Sweden (first unit mid-2030s), extending the nuclear option, and Morgan Stanley ($1,350) and Guggenheim ($1,450) raised targets. But the stock was hit hard by the AI-slowdown selloff, -8.6% on Sep 14 to $874.76 (-8% vs Aug 27, ~27% below the high), and GLJ Research initiated at Sell with a $470 target calling it 'priced for perfection'. The Oct 21 order number is the referee between those views; capacity is still scaling from 20 GW/yr to 30 GW in 2030 with pricing power rising. - HOOD:US Robinhood Markets - accumulate (medium) - Robinhood posted record Q2 2026 revenue of $1.31B (+32% YoY) with a 57% adjusted EBITDA margin, and the gap showed the prediction-markets leg maturing into real infrastructure: on Sep 8 it selected OG.com as its CFTC-regulated infrastructure partner for event contracts, will add yes/no contracts from Crypto.com's prediction business, and holds equity stakes in both partners (OG.com at a recent $20B valuation). Morgan Stanley upgraded to Overweight citing prediction markets and trading activity increasingly independent of crypto, the firm won an underwriting role on the Oura IPO, and Tenev presented at Goldman's Communacopia on Sep 9. The stock closed Sep 14 at $114.33, +1.6% on a day the AI complex fell apart - a useful demonstration that it is not an AI-capex proxy - and is +3% vs Aug 27 with consensus near $127. - MELI:US MercadoLibre - accumulate (medium) - MercadoLibre is Latin America's dominant e-commerce, logistics and fintech platform: Q2 2026 revenue grew 49.8% to $10.17B (30th straight quarter above 30%), unique buyers rose to 89M, and the Mercado Pago credit book grew 75% to $16.4B. The gap was quiet and constructive: it issued $1B of 10-year 5.850% senior unsecured notes on Sep 10 with investment-grade ratings (BBB-/Baa3) and demand from over 100 institutional investors - cheap dry powder for the logistics/credit push - and presented at Goldman's Communacopia on Sep 8. The stock sat out the AI drama entirely (+0.15% on Sep 14, closing $1,900.24, -2% vs Aug 27) and remains ~25% below its 52-week high with consensus near $2,265; earnings moved to Oct 28. One of the few names on the list whose quarter does not hinge on AI capex sentiment. - NET:US Cloudflare - buy (high) - Cloudflare's growth is re-accelerating on AI: Q2 2026 revenue rose 36% to $696M, net retention hit 120%, and its edge network is the natural control point for agentic and bot traffic. The gap stacked up proof: Adaptive Intelligence, a real-time detection engine that 'learns from trillions of daily requests', launched Aug 31; Cursor Cloud Agents now run on Cloudflare Sandboxes in customer-controlled environments (Sep 2); and on Sep 3 it partnered with OpenAI to launch an AI-driven vulnerability discovery and remediation service built on GPT-5.6 Cyber models. The payoff came on the market's worst AI day: while chips fell 5-13% on Sep 14, cybersecurity rallied and NET rose 7.8% to $330.36, pennies from its 52-week high of $334.60 and +9% vs Aug 27. Cloudflare earns from AI-agent traffic whichever lab wins - and benefits when AI anxiety pushes budgets toward security. FedRAMP High opens the federal market. - SE:US Sea Limited - buy (high) - Sea grew Q2 2026 revenue 48% to $7.8B with all three engines firing: Shopee GMV +28% to $38.3B (ad revenue +70%), Monee fintech revenue +59% with an $11.1B loan book at 1.0% NPLs, and Garena still growing; management reaffirmed 25% GMV growth and $1B+ of adjusted EBITDA for the year. The gap brought no bad news, just drift: the stock slid to $108.68 (-7% vs Aug 27) as EM Internet stayed out of favor, then rose 2.3% on Sep 14 while the AI complex sold off - defensive money is starting to notice. Arete upgraded to Buy, Benchmark's $175 target stands, consensus sits near $157 (+45%), and Shopee launched a turbo delivery service in Brazil to take the fight to Amazon and MELI. Still ~44% below the 52-week high: the widest value gap on the list for a business compounding this fast. - 1211:HK BYD - accumulate (medium) - Exports are now the growth engine: overseas sales up 134.5% YoY in August and about 44% of first-half volume, offsetting a shrinking China business. Shares near the 52-week low at HK$78.10. - 6861:JP Keyence Corp - buy (high) - Keyence is the picks-and-shovels supplier of machine vision, sensors and measurement for factory automation and the emerging humanoid/robotics supply chain; fiscal Q1 sales rose 33% to JPY 347B and net income 51% to JPY 139B with overseas sales +39%, operating margins above 50% and a 95% equity ratio. No company news over the gap, but the sell side got louder: Citi upgraded to Buy with a JPY 95,000 target, joining Goldman (JPY 100,000, APAC Conviction List), with consensus near JPY 97,900. The shares were dragged down ~7% with Japanese tech as the AI-slowdown call hit Tokyo (SoftBank -13% on Sep 14) and traded around JPY 76,270 on Tuesday morning Sep 15 (~$493, -5% vs Aug 27 in USD), now ~22% below the record. A quality compounder getting cheaper on someone else's news; October results are the next catalyst. - 7203:JP Toyota Motor - watch (medium) - The hybrid winner as EV adoption slows: hybrids are over 18% of US new vehicle sales, and the stock trades on 8.6x earnings near the low end of its 52-week range. - ASML:US ASML Holding - buy (high) - ASML is the sole supplier of EUV lithography, guiding 2026 to EUR 43-45B of sales at 54-56% gross margin, with EUV capacity for 2027 essentially fully booked and significant 2028 orders in hand. The gap added to the capacity story - the company is examining ways to build more than 110 EUV tools in 2028, is expanding in Eindhoven, and deepened advanced-node work with Samsung and Intel - but the ADR was caught square in the Sep 14 AI-slowdown selloff, falling 7.3% to $1,575.15 (-9% vs Aug 27) as equipment names took the worst of the 5.2% semiconductor-index drop. The multi-year booking visibility is exactly the defense against a sentiment-driven pause: even if hyperscalers breathe, 2027-28 EUV slots are contracted. Consensus target ~$2,160; Oct 14 results and bookings are the next data point. - TER:US Teradyne - accumulate (medium) - Teradyne is the leading supplier of automated test equipment for AI compute and HBM silicon and, via Universal Robots and MiR, the largest Western pure-play in collaborative robots; Q2 2026 revenue more than doubled to $1.33B with AI-related demand ~70% of revenue, and it joined the Nasdaq-100 in the gap. The fundamentals and the tape diverged violently: on the same Sep 14 that Universal Robots unveiled Gen 7 - its new physical-AI platform at IMTS Chicago with the PolyScope X OS, a CB7 controller packing 40% more compute in a 30% smaller footprint, and new g-Series arms - the stock fell 13.3% to $329.20, the worst one-day hit on the list, as the AI-slowdown call crushed high-multiple test names. Now -11% vs Aug 27 and ~34% below the high with consensus near $446. Nothing company-specific broke; 70% AI exposure simply cuts both ways, and the late-October print is the next chance for numbers to reassert themselves. - TSLA:US Tesla - watch (medium) - Auto demand is rolling over while the valuation leans on robotaxi, energy storage and Optimus. Track it, do not chase it, until the non-auto lines carry real revenue. ## New news (3) - 2026-09-27 [neutral] TSLA: Tesla Q3 deliveries land 2 October with estimates spread 40k wide - Consensus sits at 454,000 against Goldman at 435,000 and Barclays at 475,000, an unusually wide spread going into the print. <https://www.fool.com/investing/2026/09/27/tesla-reports-q3-deliveries-in-3-weeks-heres-the-n/> - 2026-09-17 [positive] 1211: BYD August NEV sales up 18% as overseas volume jumps 134.5% - 440,293 units sold in August, up 5% on July. Exports carried it: domestic China sales fell 14.3% YoY. <https://www.fool.com/investing/2026/09/17/byd-just-reported-record-monthly-sales-heres-what/> - 2026-09-25 [neutral] 7203: Toyota trades at 8.6x earnings near the bottom of its 52-week range - Y2,989.50 against a 52-week range of Y2,686 to Y4,000, market cap Y35.3T, as hybrids take share from pure EVs. <https://stockanalysis.com/quote/tyo/7203/> ## Notes Owner request: add Tesla, BYD and Toyota to the watchlist with buy/trim indicators. All three filed under Machinery. Watchlist cap raised from 15 to 18 in merge.py so these additions do not evict researched picks. Prices: TSLA $372.59 (27 Sep). BYD HK$78.10 (25 Sep) converted at the ~7.8 HKD peg. Toyota Y2,989.50 (25 Sep) converted at USD/JPY 157.23 (25 Sep). No holdings takes refreshed in this run - prices there are still from the 15 Sep research run.
Nothing here is financial advice. Holdings and opinions are my own; the "agent" sections are written by an automated research skill and may be wrong. Built 2026-09-28 21:35 UTC.