Zephyr — sniff.trade

Zephyr

66 episodes · 100 trades · 1 speakers

all speakers ▾

TRACKED TRADES

@zephyr_z9

4w

LTAs for B300s are between $1.8-$3/hr https://t.co/Ny33fQJ2KV

IDEAS · 3

AMZAMZNLONGAWS B300 GPU price hike of 20% on Jul 1 lifts per-GW revenue from $18.9B to $22.7B at reserved-instance rates, expanding AWS margins with zero new capex on legacy A100s$265.62+12.52%›NVDNVDALONGLTA contracts for B300 Blackwell GPUs priced at $1.8–$3/hr confirm sustained demand; AWS's 20% on-demand hike validates NVDA's chip pricing power and accelerates datacenter revenue$199.53+4.58%›GOOGOOGLLONGSector-wide GPU repricing at 20% lifts unlevered IRRs on AI datacenter capex above 20% for all hyperscalers, making Google Cloud's B300 buildout more profitable than prior models assumed$360.53+9.31%›

@zephyr_z9

4w

16 Hi HBM ain't coming anytime soon Not enough supply & lots of yield issues https:…

IDEAS · 2

MUMULONG16-layer HBM supply crunch + no challenger relief until 2028 = sustained DRAM pricing power for Micron$812.43-29.34%›SKHSKHXLONGSK Hynix leads Nvidia's HBM supply chain; 16-layer yield crisis locks in its moat for years—IDEA›

@zephyr_z9

4w

Check 3D DRAM, and u can scale till 1b node without EUV (HBM3E) https://t.co/BRZJBtDTkC

IDEAS · 2

MUMULONG3D DRAM lets HBM3E scale to the 1-billion node without EUV, removing a key cost ceiling for Micron's dominant HBM franchise$812.43-29.39%›LRCLRCXLONG3D DRAM shifts memory scaling from EUV lithography to HAR etch and ALD deposition, routing billions in capex to Lam Research's core tools$294.00-21.35%›

@zephyr_z9

4w

🫡🫡🫡🫡🫡🫡 https://t.co/JzVo6iR6BG

IDEAS · 1

EWTEWTLONGGlass fabric supercycle + foreign institutional buying surge drives Taiwan equity re-rating higher post-July$96.77-3.54%›

@zephyr_z9

4w

Memory bandwidth matters for decode and determines ur revenue https://t.co/riBsBP6ecw

IDEAS · 1

MUMULONGInference is memory-bottlenecked, making Micron the direct revenue lever for AI inference at scale$812.43-29.46%›

@zephyr_z9

4w

Mustafa finally getting sidelined?? https://t.co/ge7WzbMxDn

IDEAS · 1

MSFMSFTLONGMSFT's Copilot overhaul replaces a struggling AI lead with a consumer-growth specialist, signaling urgency to arrest a market share collapse and retake the AI product race.$471.28+28.73%›

@zephyr_z9

4w

$ACMR and Naura $002371 might receive some orders, any other potential beneficiaries (a…

IDEAS · 2

ACMACMRLONGACMR named as direct WFE order beneficiary; 65% Q1 order growth at ACM Shanghai signals accelerating 2026 revenue$76.80-25.33%›MKSMKSILONGWFE subsystem suppliers are the deeper beneficiary of China capex surge; MKSI provides components to every tool maker in the chain$299.96-24.02%›

@zephyr_z9

4w

Good point CXMT will generate over $130B in gross profit in the next 18 months How the…

IDEAS · 2

MUMULONGCXMT's $130B+ gross profit over 18 months signals a surging memory upcycle, lifting all DRAM prices and Micron's earnings$812.43-28.10%›LRCLRCXLONGCXMT's 100B+ RMB profit reinvested as capex into domestic semiconductor equipment surges orders for memory-focused WFE makers like Lam Research$294.00-23.72%›

@zephyr_z9

4w

Good first gen product But 280mW and 24 fibers won't cut it https://t.co/3QZRPbhHzQ

IDEAS · 1

COHCOHRLONGGlassBridge demo specs fall short of real CPO demands; Coherent is the verified silicon photonics winner with NVIDIA locked in through 2030.$258.28-33.71%›

@zephyr_z9

4w

Has been in production since early May https://t.co/brz3W0sRQj

IDEAS · 1

NVDNVDASHORTDeepSeek's DSpark slashes inference cost ~5x, eroding the GPU demand narrative that props up Nvidia's valuation$199.53-3.17%›

@zephyr_z9

4w

CXMT will probably list at $500B (some are predicting $700B) on the Shanghai Star Marke…

IDEAS · 1

MUMULONGCXMT's $500-700B IPO validates the DRAM supercycle; Micron is the most liquid Western pure-play on the same trade$812.43-29.14%›

@zephyr_z9

4w

The problem isn't just the price Getting enough DRAM is hard Secondly, this is China on…

IDEAS · 1

AAPAAPLSHORTApple's China smartphone share is structurally eroding as foreign phone shipments fell 19.2% YoY while Huawei gains with domestic chip advantage.$303.45-4.83%›

@zephyr_z9

4w

Very important paradigm Rn all the computer use AI models are janky & slow Small mo…

IDEAS · 2

AMDAMDLONGLow-latency AI inference chips are the critical bottleneck for computer-use agents; AMD captures the CPU demand surge as agentic AI scales$482.97-7.11%›AMZAMZNLONGAnthropic's aggressive computer-use AI expansion compounds Amazon's existing $13B Anthropic bet and a $100B AWS commitment, making AMZN the primary public beneficiary$265.62+15.62%›

@zephyr_z9

4w

This pretty much confirms that KYC for AI model subscription is coming soon https://t.c…

IDEAS · 1

MSFMSFTLONGUS government gates frontier AI access to approved users, making OpenAI/MSFT the default winners of a restricted market$471.28+26.18%›

@zephyr_z9

4w

ASIC suppliers are having a very hard time securing HBM supply There are two 800-pound …

IDEAS · 3

NVDNVDALONGNVIDIA controls HBM supply allocation at scale, locking out ASIC rivals and compounding its AI chip pricing power$199.53+3.15%›AVGAVGOLONGBroadcom's Hock Tan is the second HBM gatekeeper, cementing AVGO's custom AI chip moat across Google, Meta, and OpenAI$379.98+5.29%›MUMULONGHBM scarcity is so extreme even Google can't source it directly, making Micron's constrained memory supply a multi-year pricing windfall$812.43-28.28%›

@zephyr_z9

4w

Only $194B or $19.4B/yr on chip-making facilities Are they serious?? https://t.co/DAWed…

IDEAS · 2

SMSSMSNLONGSamsung's $648B investment package — $194B directly into chip fabs — is the largest semiconductor capex commitment in history, bullish for the stock.—IDEA›SKHSKHXLONGSouth Korea's AI chip buildout is centered on two companies; Samsung's $194B fab expansion validates the entire Korean memory ecosystem, including SK Hynix.—IDEA›

@zephyr_z9

1mo

Contact | Citrini Research

IDEAS · 1

MRVMRVLLONGMarvell's deepening TPU content with Google and Nvidia validates bullish analyst targets; custom silicon revenue set to more than double by FY2028.$184.60-31.80%›

@zephyr_z9

1mo

HUH 5.6 is a big boi model https://t.co/gvubgJnEvb

IDEAS · 1

CBRCBRSLONGOpenAI deploying GPT-5.6 Sol on Cerebras hardware at 750 tokens/sec validates Cerebras as OpenAI's frontier inference partner, reinforcing a $20B+ contract backlog.$200.42+12.37%›

@zephyr_z9

1mo

Everyone is adopting memory on logic 3D stacking Samsung, Huawei, Apple, Qualcomm for p…

IDEAS · 4

QCOQCOMLONGQualcomm's HBC 3D stacking architecture opens a $15B data center revenue stream by 2029, re-rating it far beyond a mobile-chip company.$149.12-19.71%›NVDNVDALONGNvidia's confirmed 3D logic stacking in its 2028 Feynman GPU generation adds another step-change in AI compute performance, sustaining its hardware lead.$199.53+1.42%›TSMTSMLONGEvery named company adopting 3D stacking must go through TSMC's SoIC packaging, making it the unavoidable toll booth of the memory-on-logic wave.$394.96-8.88%›AVGAVGOLONGBroadcom's 3D stacking adoption for AI XPUs by 2028-29 accelerates its custom silicon franchise with Google, OpenAI, Meta, and ByteDance as anchor customers.$379.98+4.32%›

@zephyr_z9

1mo

CXMT targeting over 500Mb/mm2 with 64-layer 3D DRAM test vehicle | SemiWiki

IDEAS · 1

MUMULONGCXMT's 3D DRAM milestone validates the memory scaling supercycle; Micron leads execution with HBM sold out and blowout earnings$812.43-27.05%›

@zephyr_z9

1mo

Can someone supply me the hybrid bonded HBM4?? "HBM4 leans on hybrid bonding even hard…

IDEAS · 1

ADEADEALONGAdeia's flat-fee HBM memory licenses expire 2027-2028 exactly as HBM4 ramps, giving them rare leverage to convert to per-chip royalties and re-rate the stock.$26.95-12.20%›

@zephyr_z9

1mo

WOW https://t.co/BwENdakeK6

IDEAS · 2

SYNSYNALONGonsemi's all-stock acquisition of SYNA at 1.350 ON shares per share locks in a deal premium, pulling SYNA toward implied value.$108.56-11.95%›ONONSHORTonsemi issues ~12% dilution to fund the SYNA acquisition; acquirer stock typically sells off on deal announcement.$81.24+10.59%›

@zephyr_z9

1mo

The US Government has requested a slow staggered rollout of GPT-5.6, and OpenAI has agr…

IDEAS · 1

MSFMSFTSHORTGovernment-mandated customer-by-customer GPT-5.6 approval will strangle OpenAI ARR growth, hitting Azure OpenAI revenue at Microsoft$471.28-25.09%›

@zephyr_z9

1mo

DSV4-Flash is loved. Gemini 3 Flash is loved. The first Haiku was deemed blessed. o4-mi…

IDEAS · 1

GOOGOOGLSHORTCheap small models cannibalize flagship AI revenue; Alphabet's topline under pressure as Flash/mini erode premium pricing power$360.53-7.30%›

@zephyr_z9

1mo

No memory anon https://t.co/9FHV8OK5ct

IDEAS · 1

AAPAAPLLONGApple's AI-focused Mac silicon roadmap (M7 Pro/Max 2027, M7 Ultra 2028) sets up a multi-year upgrade supercycle.$303.45+8.40%›

@zephyr_z9

1mo

Late 2025 and Q1 2026 stockpiles are over Now, we are going to get a lot of price hikes…

IDEAS · 1

AAPAAPLLONGApple's global hardware price hikes expand ASPs and margins as stockpile front-loading ends, making AAPL a buy.$303.45+5.31%›

@zephyr_z9

1mo

Must read for everyone I agree with most points written in the article Although, I disa…

IDEAS · 4

SKHSKHXLONGHBM enters a multi-year structural supercycle with 2.5/3 escape conditions met; SK Hynix wins the quality-competition race—IDEA›MUMULONGCommodity DRAM enters a structural supercycle as agentic AI drives ~50% demand CAGR vs ~20% supply CAGR, shortage gap widens through 2030$812.43-32.12%›SNDSNDKLONGNAND SSD already in a worse shortage than DRAM; KV cache offloading, AI video, and agent sandbox demand create structural supercycle through 2030$1,235.36-47.76%›AMDAMDLONGCPU server TAM forecasts being rapidly doubled by AMD, Nvidia, and Bernstein as agentic AI shifts CPU:GPU ratios and creates a new structural demand wave$482.97-11.19%›

@zephyr_z9

1mo

The only caveat here is that they won't use HCB till 2028 https://t.co/geCISiDs22

IDEAS · 1

SKHSKHXLONGSamsung's HCB deployment delayed to 2028, leaving SK Hynix dominant in HBM through the AI buildout—IDEA›

@zephyr_z9

1mo

Nice https://t.co/njC6tu2X1R

IDEAS · 1

MUMULONGMicron's SCAs locking ~40% of revenue at CQ2 price floors turns volatile memory into predictable infrastructure revenue, re-rates the stock higher$812.43-32.55%›

@zephyr_z9

1mo

Read carefully Memory FUD bois https://t.co/XKm7ZMcbHz

IDEAS · 1

MUMULONGMicron CEO confirms no visibility on when memory supply catches demand, locking in durable pricing power for MU.$812.43-31.48%›

@zephyr_z9

1mo

Where are the memory FUD bois??? https://t.co/yttDA5QWDM

IDEAS · 1

MUMULONGMicron crushes Q3 estimates on revenue guide and margin, demolishing the memory bear case$812.43-26.91%›

@zephyr_z9

1mo

Deep|LLM: 26H1 Update (Part 2): Frontier Labs, Chinese Model Vendors & the Compute Bott…

IDEAS · 3

NVDNVDALONGH100 lease rates +40% since Oct 2025 confirm GPU compute is the binding bottleneck of AI infrastructure, pricing power flows to NVDA$199.53-0.67%›AMZAMZNLONGAnthropic ARR exploded from $1B to $47B+ in 18 months; Amazon's $13B+ stake and AWS as the delivery cloud make AMZN the best public proxy$265.62+12.25%›AVGAVGOLONGAgentic AI driving a custom ASIC supercycle; Broadcom is the co-designer of record for OpenAI and Google, with AI revenue set to double to $42B by 2027$379.98+0.26%›

@zephyr_z9

1mo

70% GM and 15%-20% FCF margin Damn Anthropic is printing cash now https://t.co/f4Ce8aBT…

IDEAS · 2

NVDNVDALONGAnthropic's GPU hunger and H100 lease rate rebound confirm physical compute is the binding constraint, driving NVDA higher$199.53-1.11%›AMZAMZNLONGAmazon's $13B Anthropic stake is now worth multiples more as Anthropic hits $47B ARR and files for IPO, while AWS captures every dollar of compute spend$265.62+12.70%›

@zephyr_z9

1mo

Another way to scale revenue is to sign deployment agreements with hyperscalers and inf…

IDEAS · 2

AMZAMZNLONGKimi on AWS Marketplace deepens AWS's role as the default inference layer for third-party AI models, expanding its AI revenue flywheel$265.62+12.16%›MINMINIMAXLONGChinese AI foundation model companies are cracking global distribution via hyperscaler marketplaces, opening a path to $5B–$10B ARR that the market has not priced in—IDEA›

@zephyr_z9

1mo

Very interesting https://t.co/7bc0XfzqTN

IDEAS · 1

AMZAMZNLONGKimi API joins AWS Marketplace with EDP billing integration, deepening AWS's third-party AI model flywheel and enterprise cloud consumption.$265.62+12.55%›

@zephyr_z9

1mo

I don't think it will pass cuz it will shut down a large chunk (25%-30%) of global DRAM…

IDEAS · 2

SKHSKHXSHORTMATCH Act in NDAA FY27 forces closure of SK Hynix China fabs, wiping 25-30% of global DRAM/NAND supply and crushing its stock—IDEA›ASMASMLSHORTMATCH Act cuts ASML off from 29% of its revenue by blocking tool sales and servicing to Chinese fabs$1,644.70+8.72%›

@zephyr_z9

1mo

IDEAS · 1

AMZAMZNLONGAnthropic IPO unlocked by Trump admin thaw; Amazon's $13B+ stake gets a direct valuation catalyst.$265.62+11.68%›

@zephyr_z9

1mo

Looks like the next China-US export control showdown is near MATCH Act has been added t…

IDEAS · 2

ASMASMLSHORTMATCH Act in NDAA FY27 threatens ASML's ~29% China revenue as full DUV blockade advances toward House floor vote in July.$1,644.70+7.56%›MUMULONGMATCH Act cuts off Chinese memory rivals CXMT and YMTC from equipment and servicing, entrenching Micron's dominance in DRAM and HBM.$812.43-21.36%›

@zephyr_z9

1mo

> According to the official document, a 1-second seedance video requires 20,000 for …

IDEAS · 1

NVDNVDALONGAI video gen at planetary scale implies quadrillions of tokens/day, requiring far more GPU compute than currently exists$199.53+0.27%›

@zephyr_z9

1mo

OpenAI loves TPUs (or their designs) https://t.co/ewriQj58gb

IDEAS · 3

AVGAVGOLONGJalapeño is a confirmed multi-year design win for Broadcom, deepening its custom ASIC franchise with OpenAI.$379.98+0.48%›NVDNVDASHORTOpenAI's custom inference chip reduces GPU demand from the largest AI buyer, pressuring Nvidia's inference pricing power.$199.53-1.95%›CLSCLSLONGCelestica is the exclusive system integrator for OpenAI's Jalapeño racks, securing a multi-year hardware build-out contract.$337.09-10.02%›

@zephyr_z9

1mo

> According to the official document, a 1-second seedance video requires 20,000 for …

IDEAS · 1

NVDNVDALONGMultimodal token explosion at ByteDance (65Q tokens/yr) validates GPU demand far exceeding consensus$199.53-2.98%›

@zephyr_z9

1mo

Pumper in chief will keep us entertained for the next 15 years Long live Masa https://t…

IDEAS · 1

SFTSFTBYLONGMasa Son staying at SoftBank 10-15 more years keeps bold tech bets alive, sustaining the hype cycle that drives SoftBank's valuation premium.$15.31-25.96%›

@zephyr_z9

1mo

LMAO Who the hell did the ASP math at BoFA Contract prices will at least hit $30/GB nex…

IDEAS · 1

MUMULONGBofA's HBM ASP math is wrong; contract prices hitting $30/GB next year drives Micron's earnings power sharply higher$812.43-25.00%›

@zephyr_z9

1mo

More capital for model development https://t.co/SNSKBKc9n4

IDEAS · 1

ZHIZHIPULONGZhipu is tapping markets again for billions in fresh capital, doubling down on model R&D just as GLM-5.2 sends the stock to new highs.—IDEA›

@zephyr_z9

1mo

ICBC generates around $60B in net income CXMT will generate over $80B next year Will be…

IDEAS · 2

SKHSKHXLONGCXMT's DRAM supercycle profits signal a rising tide for all major DRAM makers, including SK Hynix—IDEA›MUMULONGMicron earnings on June 24 should confirm the same DRAM supercycle crushing it for CXMT, with Wall Street expecting $35B revenue and $20.57 adj. EPS$812.43-22.37%›

@zephyr_z9

1mo

DAYUM CXMT doubling revenue to $100B-$110B in 2027 CRAZY https://t.co/IgluhQwYkV https:…

IDEAS · 2

MUMULONGCXMT's $55B 2026 DRAM revenues confirm a supercycle that directly boosts Micron, its closest liquid analog$812.43-25.85%›DRADRAMLONGCXMT's $55B 2026 DRAM revenues confirm AI memory supercycle powering the entire sector; DRAM ETF captures the basket$51.01-27.04%›

@zephyr_z9

1mo

GPU Tsunami Beneficiaries: Power Semis and Analog

IDEAS · 2

MPWMPWRLONGAI rack density scaling pulls power semiconductor content onto the tray itself, making MPWR the dominant pick-and-shovel play on the GPU power delivery bottleneck.$1,427.63-0.83%›TXNTXNLONGAnalog control chips — sensing, regulation, eFuse, PMICs, telemetry — are a mandatory and growing layer in every AI rack, and TXN's data center revenue is accelerating at 90% YoY to confirm the structural shift.$274.61-10.70%›

@zephyr_z9

1mo

China’s CXMT Is Set to Challenge DRAM Incumbents

IDEAS · 1

MUMULONGA once-in-four-decades DRAM shortage is driving explosive ASP growth, and Micron is the cleanest public route to the supercycle.$812.43-25.48%›

@zephyr_z9

1mo

China’s CXMT Is Set to Challenge DRAM Incumbents

IDEAS · 2

SMSSMSNLONGDRAM ASPs on track to double in 2026 — a once-in-four-decades shortage lifts Samsung's memory revenue to record highs—IDEA›MUMULONGDRAM ASPs doubling in 2026 drives Micron revenue to record highs, with Q3 earnings in two days as the next proof point$812.43-26.45%›

@zephyr_z9

1mo

China’s CXMT Is Set to Challenge DRAM Incumbents

IDEAS · 2

MUMULONGDRAM ASP supercycle is doubling prices industry-wide, driving explosive revenue and margin expansion at Micron$812.43-25.50%›SKHSKHXLONGDRAM ASP supercycle lifts SK Hynix revenue toward ~$56B quarterly by 4Q26E as industry-wide pricing doubles—IDEA›

@zephyr_z9

1mo

Vanadium redox batteries for DC electricity storage This Xinjiang project (1.93 Yuan/Wh…

IDEAS · 1

LGOLGOLONGChina's largest VRFB project reaches commercial parity with LFP, signaling rising vanadium demand that directly lifts the world's leading pure-play vanadium miner$0.58-19.49%›

@zephyr_z9

1mo

Harlan hardest hit https://t.co/uzH2LJ1fB9

IDEAS · 1

AVGAVGOSHORTGoogle's Pumafish cancellation (Broadcom's TPU v9 design win) removes a key AI revenue leg as MediaTek locks up Humufish and Triggerfish exclusively$379.98-0.74%›

@zephyr_z9

1mo

LOL Expect large upward revisions in HBM contract prices Memory trio will milk it https…

IDEAS · 2

MUMULONGMemory trio throttles HBM supply → large upward revisions in HBM contract prices → Micron earnings beat incoming$812.43-32.22%›SKHSKHXLONGSK Hynix pivots capacity from HBM4 to commodity DRAM, capturing near-90% operating margins as supply shortages peak—IDEA›

@zephyr_z9

1mo

"Marvell which you would think should be competing is just a pump and dump by the manag…

IDEAS · 2

ALAALABLONGAstera Labs is the only end-to-end CXL stack play; switches + software + cards compound into durable moat as the $20B scale-up market matures.$310.05-28.51%›MRVMRVLSHORTMarvell is mis-priced as a premium ASIC designer when its ASIC business is commodity IT outsourcing; management extraction makes the premium unjustifiable.$184.60+39.55%›

@zephyr_z9

1mo

https://t.co/nglm4ajymc https://t.co/GtzQBXymeP

IDEAS · 1

MUMULONGLegacy DRAM prices up 55-60% QoQ with another 35-40% ahead; Micron as top DRAM producer takes the most upside$812.43-34.97%›

@zephyr_z9

1mo

YUP https://t.co/wJMJjLnjNl

IDEAS · 1

AMZAMZNSHORTFable ban kills Anthropic's rev acceleration; AWS/Bedrock is the distribution pipe that goes dark, hurting AMZN$265.62-15.85%›

@zephyr_z9

1mo

MSFT will kill a lot of loss-making units to fund the CAPEX https://t.co/LD3sdn7GLm

IDEAS · 1

MSFMSFTLONGMicrosoft axing loss-making Xbox units frees capital for $190B AI CAPEX, unlocking margin expansion and cloud upside.$471.28+25.95%›

@zephyr_z9

1mo

This transformation will happen within 5 years https://t.co/3GNOUELSAu

IDEAS · 1

JDJDLONGJD.com eliminates 700k delivery worker salaries via robots, compressing its largest cost line within 5 years.$32.23+16.20%›

@zephyr_z9

1mo

Hmmm... What's up with the crazy jump in HBM exports?? HBM4 started shipping for Rubin …

IDEAS · 2

SKHSKHXLONGHBM4 mass shipments to Nvidia Rubin MP have started, and SK Hynix as the dominant HBM4 supplier captures the revenue surge now showing in Korean export data.—IDEA›NVDNVDALONGHBM4 now shipping to Nvidia's Rubin MP confirms the next-gen AI GPU supply chain is executing, removing a key risk to Nvidia's forward revenue ramp.$199.53-4.79%›

@zephyr_z9

1mo

Any company associated with CXL will go bananas

IDEAS · 1

ALAALABLONGCXL adoption boom makes Astera Labs — the only pure-play public CXL company — the biggest beneficiary as AI data centers scale memory interconnects.$310.05-25.48%›

@zephyr_z9

1mo

ALD https://t.co/XSGcjDpdeV

IDEAS · 1

ASMASMIYLONGASM International is the world's leading pure-play ALD equipment supplier, up 92.58% YTD as chipmakers race to deploy atomic layer deposition at advanced nodes.$937.09-24.00%›

@zephyr_z9

1mo

It's not impossible to match Mythos All u need is a solid 3T-4T parameter base model an…

IDEAS · 1

MINMINIMAXLONGZhipu has the post-training capability to match Anthropic's Mythos, lifting the ceiling for Chinese AI foundation model peers like MiniMax.—IDEA›

@zephyr_z9

1mo

Anthropic is beefing up its Biology/Chemistry AI team https://t.co/UT1wXnqmnL

IDEAS · 1

GOOGOOGLSHORTNobel Prize-winning AlphaFold lead defecting from DeepMind to Anthropic signals Google's AI science bench is weakening.$360.53+1.74%›

@zephyr_z9

1mo

Things are not looking good for Google They have missed the agentic coding train Multim…

IDEAS · 2

GOOGOOGLSHORTGoogle has fallen to 5th place on AI benchmarks with talent fleeing and no competitive frontier model, threatening its core tech premium.$360.53+1.99%›MSFMSFTLONGMicrosoft owns GitHub Copilot and the OpenAI partnership, making it the direct winner of the agentic coding wave Google missed.$471.28+26.93%›

@zephyr_z9

1mo

China will be huge for BESI as Huawei is adopting HB for Logic Folding U can even see i…

IDEAS · 1

AMAAMATLONGHuawei's LogicFolding and CPO adoption drive accelerating hybrid bonding equipment demand, with AMAT co-leading the buildout alongside BESI.$504.12-17.44%›

@zephyr_z9

1mo

They seem to be focused on coding rn I'm pretty sure this is the largest coding trainin…

IDEAS · 1

METMETALONGMeta is running the world's largest coding AI training pipeline; this compute and talent bet is a long-term moat for META.$565.55-2.12%›