Plutux

Semiconductors

Chips, from the fab floor to the income statement

Foundry capacity, HBM supply, packaging bottlenecks and export rules — traced through to the companies whose quarters they decide.

What Big Funds Own22 fundsWhat Berkshire, Bridgewater and other big funds reported holding last quarter.What Analysts Expect16 stocksRatings and price targets from the big institutions, each with its public source.

2026-09-07

The Reported $20B SK Hynix FX Trade Turns AI Memory Into Korea’s Won Shock Absorber insight cover
Markets / Event
SKHY•17 min read

The Reported $20B SK Hynix FX Trade Turns AI Memory Into Korea’s Won Shock Absorber

Reuters reported that South Korea’s foreign-exchange authorities bought about $20 billion of dollars sold by SK Hynix after its July Nasdaq offering, while official government disclosures confirm that the ADR inflow helped move the won from the 1,550 range toward the upper 1,400s. The important distinction is that the public record supports a state absorption trade, not a confirmed legal order forcing repatriation. AI-memory profits made the flow large enough to matter for currency policy, but the intervention also makes SK Hynix’s dollar-conversion choices more policy-sensitive.

Reported FX-fund purchase: $20BReported ADR proceeds: $26.5B
Read article

2026-09-06

Nvidia Closed Green on the Day the Jobs Beat Sank Everything Else — the AI Capex Trade Has Stopped Waiting for the Fed insight cover
Markets / Event
NVDA · 000660.KS•15 min read

Nvidia Closed Green on the Day the Jobs Beat Sank Everything Else — the AI Capex Trade Has Stopped Waiting for the Fed

On Friday Sept 4, 2026, NVIDIA closed at $230.36 — within $5.38 of its $235.74 record — while the S&P 500 (-0.38%), Nasdaq (-0.29%) and Dow (-0.51%) all fell after a hawkish August jobs beat (+162k vs. 56k consensus) pushed the 2-year Treasury yield 5bps higher to 4.38%. The split is the cleanest read yet that AI infrastructure capex has become its own cycle: hyperscalers are committing more than $600B in 2026 alone, SK Hynix-Nvidia have a $500B+ initiative on the books, and Micron's 2026 HBM is fully sold out. The bottleneck is supply, not demand — and it sits above the foundry, in TSMC CoWoS packaging and HBM memory, not at the GPU itself.

NVIDIA Sept 4, 2026 close: $230.36Aug nonfarm payrolls: +162,000
Read article
TSMC's 20-fab buildout shows the AI-demand worry is obsolete — equipment is the bottleneck insight cover
Supply Chain
TSM · NVDA•14 min read

TSMC's 20-fab buildout shows the AI-demand worry is obsolete — equipment is the bottleneck

TSMC is constructing 13 fabs in Taiwan and 5-6 overseas simultaneously, with 2026 capex lifted to $60-64B and equipment procurement demand up 90% in eight months. Q2 2026 gross margins hit a record 67.7% and NVIDIA has locked in 60% of CoWoS output, so the binding constraint has flipped from demand to ASML, Tokyo Electron and the rest of the equipment chain. Investors should size positions to the equipment suppliers and the customer anchors, not the foundry itself.

Fabs under simultaneous construction: ~202026 capex guidance (midpoint): $62B
Read article

2026-09-05

Ambarella's print passed the test. The tape still doesn't trust edge-AI silicon. insight cover
Earnings
AMBA · NXPI•12 min read

Ambarella's print passed the test. The tape still doesn't trust edge-AI silicon.

Ambarella posted Q2 FY2027 revenue of $108.1M (+13.2% YoY) and non-GAAP EPS of $0.18, both ahead of consensus, with a record edge-AI quarter driven by 5nm CV75/CV72 SoCs and a Q3 guide of $115–124M that would push sequential growth up to ~15%. The market response — an ~11% post-print slide and four broker downgrades the next morning — undercuts the headline, pinning the worry on supply-chain costs and a flat H2 auto ramp. With NXP Semiconductors circling at roughly $3.3B, Samsung Foundry at full tilt, and Hanwha Vision locked in under an $800M-plus decade-long agreement, the silicon story is real; the multiple is what is being repriced.

Q2 FY2027 revenue: $108.1MQ3 FY2027 revenue guide: $115–124M
Read article
Foxconn's Q3 call turns the AI-server read from 'strong demand' into '2027 capacity stays very strong' insight cover
Industry News
DELL · AVGO•15 min read

Foxconn's Q3 call turns the AI-server read from 'strong demand' into '2027 capacity stays very strong'

Foxconn's August 12 earnings did more than beat Q2 — management explicitly told investors that 2027 demand for AI production capacity will stay very strong as Vera Rubin racks enter mass production this quarter, with cloud & networking guided up 'high double-digit' both QoQ and YoY. That forward signal lands alongside a cross-validating cluster: Dell's record $95B AI backlog, Broadcom's $21.7B Q4 AI-semi guide (+236% YoY), NVIDIA's $89B data-center print (+117% YoY), Wistron's AI servers hitting 95% of revenue, and Lenovo's $54B AI-server pipeline (+157% QoQ). The structural story is sharper than consensus: AI servers just crossed 51% of Foxconn's revenue, displacing iPhone as the swing factor and reframing the stock as an AI-infrastructure bet.

Foxconn Q2 2026 revenue: NT$2.53TFoxconn Q2 2026 net profit: NT$59.97B
Read article
SanDisk Soars Into the S&P 100 — With Hedge Fund Holdings Already Doubled, the Forced-Buyer Trade Gets Its Real Test insight cover
Markets / Event
SNDK · DELL•7 min read

SanDisk Soars Into the S&P 100 — With Hedge Fund Holdings Already Doubled, the Forced-Buyer Trade Gets Its Real Test

SanDisk joins the S&P 100 on September 21, 2026, alongside Dell Technologies, Palo Alto Networks and Arista Networks, pulling every S&P 100 index fund into a stock that's already rallied roughly 600% year-to-date. With 128 hedge funds now holding $25.6B of shares (up 125% quarter-on-quarter) and short interest still at 5.3% of float, this isn't just another index add — it's the first structural forced-buyer test for the AI-storage trade, and the re-rating-versus-derating risk lives in the gap between passive flow and crowded positioning.

S&P 100 inclusion: Sept 21, 2026Hedge funds holding SNDK: 128 funds
Read article

2026-09-04

NVIDIA's RTX Spark PC launch in October tests whether on-device inference becomes a new revenue engine—or a cloud-demand throttle insight cover
Industry News
NVDA•8 min read

NVIDIA's RTX Spark PC launch in October tests whether on-device inference becomes a new revenue engine—or a cloud-demand throttle

NVIDIA’s RTX Spark moves “personal AI agents” onto Windows PCs designed for local inference, with OEMs signaling an October arrival window. The investor question is binary: does on-device execution expand GPU/PC demand for NVIDIA, or does it quietly reduce incremental inference demand that has been flowing to data centers and cloud partners?

NVIDIA revenue: $302.97BNVIDIA net income: $192.88B
Read article
Trump’s “chip shield” raises the price of being “American-made” — Tennessee’s polysilicon plant becomes the first real test case insight cover
Industry News
9 min read

Trump’s “chip shield” raises the price of being “American-made” — Tennessee’s polysilicon plant becomes the first real test case

A Reuters investigation describes how a Tennessee polysilicon project faces closure risk as a “shield” meant to favor U.S.-sourced inputs ends up taxing the downstream supply chain instead. The policy mechanism—tariffs plus price floors that do not actually differentiate by where the polysilicon was sourced—pushes U.S. input costs higher and helps drive away the remaining customers needed to keep the plant running.

Micron FY2024 revenue: $25.1BMicron FY2025 revenue: $37.4B
Read article

2026-09-03

Can “neutral” colocation capture AI inference profit—now that Nvidia is building inside Equinix’s fabric? insight cover
Industry News
9 min read

Can “neutral” colocation capture AI inference profit—now that Nvidia is building inside Equinix’s fabric?

Equinix’s Sept. 2, 2026 launch of “Equinix Inference Exchange,” anchored on Nvidia’s enterprise reference architecture and Together AI’s inference platform, is a bid to make vendor-neutral inference easier to deploy across Equinix’s interconnection layer. The investor question isn’t whether AI demand grows—it’s whether this packaging changes where the margin lands: colocation economics, or the hyperscalers’ walled-garden inference/distribution advantage.

TTM revenue: $9.83BTTM net income: $1.53B
Read article

2026-09-02

Broadcom beat-and-fell: the weak guide turned the custom-silicon backlog “test” negative for AI racks insight cover
Earnings
AVGO•7 min read

Broadcom beat-and-fell: the weak guide turned the custom-silicon backlog “test” negative for AI racks

Broadcom’s Q3 FY2026 results showed AI acceleration, but management’s next-quarter AI revenue guide did not provide the backlog-to-revenue “conversion proof” investors wanted. With AI semiconductor revenue guided to $21.7B for Q4 (+236% YoY), the market still sold Broadcom—signaling that the real debate is whether AI custom silicon is arriving on schedule for hyperscaler racks, not whether AI demand exists.

AI semiconductor revenue (Q3 FY2026): $16.7BAI semiconductor revenue guide (Q4 FY2026): $21.7B
Read article
Credo’s post-earnings drop tests whether copper interconnect cash flows can keep up with AI rack spend — even as Dell confirms demand is accelerating insight cover
Earnings
CRDO•7 min read

Credo’s post-earnings drop tests whether copper interconnect cash flows can keep up with AI rack spend — even as Dell confirms demand is accelerating

Credo CRDO reported Q1 FY2027 results above the optics-friendly narrative investors were pricing, but the stock sold off sharply the next day. The key question for AI rack investors is whether the market is rotating from “copper rack capture” to “optics/other bottlenecks,” and whether Credo’s guidance supports that transition.

Revenue: $479.0MGAAP gross margin: 64.5%
Read article
Taiwan’s $20B U.S. push isn’t “Made in USA”—it’s a tariff-shaped fab and factory supply-chain buildout insight cover
Supply Chain
9 min read

Taiwan’s $20B U.S. push isn’t “Made in USA”—it’s a tariff-shaped fab and factory supply-chain buildout

Taiwan’s government-linked disclosure of an additional $20B of U.S. investment reinforces a clear pattern: tariff-linked relocation is concentrating capacity buildout in Asian electronics ecosystems, not broadly “reshoring” the whole industrial base. The investable opportunity clusters upstream in tools/material handling and gases, and downstream in EMS/thermal/mechanical integration—where relocation turns into orders, not press headlines.

Taiwan–U.S. investment framework (semis/adjacent: $250BIncremental credit to facilitate investment: $250B
Read article

2026-09-01

IBM’s Aug 26 close of HRL Laboratories turns “defense-relevant” quantum sensing into an owned capability—not just a partnership insight cover
Industry News
9 min read

IBM’s Aug 26 close of HRL Laboratories turns “defense-relevant” quantum sensing into an owned capability—not just a partnership

IBM closed its acquisition of HRL Laboratories on Aug. 26, 2026, taking control of a private R&D lab jointly owned by Boeing and General Motors and focused on quantum computing and quantum sensing. The deal matters to investors because it bundles quantum hardware stack elements (including cryogenics, control electronics, packaging) that are prerequisites for turning lab-grade sensors into scalable, sovereign-capable defense-adjacent systems.

IBM Quantum Starling target: 2029Starling capability claim: 20,000×
Read article

2026-08-31

Ciena's order book is the missing systems-level proof for AI networking demand — but it comes with a demand-vs-supply timing squeeze insight cover
Earnings
CIEN•7 min read

Ciena's order book is the missing systems-level proof for AI networking demand — but it comes with a demand-vs-supply timing squeeze

Ciena’s latest reported results show strong growth alongside a large backlog, but the most decision-relevant question—whether AI demand is showing up at the optical-systems layer rather than only upstream modules—can’t be answered from the primary sources available in this research run. The numbers we can verify support a real order-book-backed cycle expansion, while key “webscale/AI” backlog breakdowns and “optical systems” attribution were not disclosed in the pages that loaded successfully.

TTM revenue: $5.57BTTM net income: $438.3M
Read article
Nvidia’s $3.5B MediaTek bet reframes “hyperscaler ASICs” into Nvidia-routed, rack-scale custom chips—raising the odds it slows the GPU exodus insight cover
Industry News
7 min read

Nvidia’s $3.5B MediaTek bet reframes “hyperscaler ASICs” into Nvidia-routed, rack-scale custom chips—raising the odds it slows the GPU exodus

Nvidia’s $3.5B investment in MediaTek is more than capital—it’s an ecosystem play: MediaTek will adopt Nvidia’s NVLink Fusion platform to help customers field custom XPUs inside Nvidia-connected AI racks. If hyperscalers increasingly buy “integration-ready” custom accelerators instead of building their own end-to-end stack, Nvidia can keep participating in training/inference demand even as silicon becomes more customized.

Nvidia revenue (TTM): $303.0BNvidia operating cash flow (TTM): $134.4B
Read article

2026-08-30

ABF substrates are the one shared bottleneck in the GPU–custom-ASIC race insight cover
Supply Chain
NVDA•8 min read

ABF substrates are the one shared bottleneck in the GPU–custom-ASIC race

Even as NVIDIA chips and Google-style custom silicon chase compute per watt, both workflows still converge on ABF build-up film and the laminate stack built from it. Ajinomoto’s ABF price move and Ibiden’s multi-year high-end substrate capex outline a critical supply-chain reality: the AI build-out is not zero-sum on compute—it is constrained by specific materials and laminate capacity.

FY2025 revenue: ¥1.53TFY2024 revenue: ¥1.44T
Read article
Marvell is down ~6% because the market treats custom silicon as a trade-off, but earnings show hyperscalers are funding both insight cover
Markets / Event
MRVL•8 min read

Marvell is down ~6% because the market treats custom silicon as a trade-off, but earnings show hyperscalers are funding both

Marvell’s Aug 2026 earnings-week narrative looked like a zero-sum fight between Nvidia GPUs and hyperscaler custom silicon. But Marvell also disclosed a Google custom-silicon framework that explicitly attaches to the TPU ecosystem, while Nvidia’s earnings reinforced that hyperscaler demand isn’t stalling. The lesson: custom silicon is additive capacity for the AI stack, not a replacement cycle—so the tape is likely mispricing who grows faster as budgets expand.

Marvell Q2 FY2027 net revenue: $2.739BMarvell TTM revenue: $9.450B
Read article
Broadcom’s September “test” hinges on whether backlog turns into revenue without breaking the $29B AI-rack backstop math insight cover
Earnings
10 min read

Broadcom’s September “test” hinges on whether backlog turns into revenue without breaking the $29B AI-rack backstop math

Broadcom’s latest SEC disclosures describe a compute-capacity/AI-rack backstop structure with a maximum exposure of $29 billion, while also reporting $164.6B in remaining performance obligations and $164.2B+ in multi-quarter contract balances. For investors, the Q3 FY2026 guide is less about growth claims and more about whether custom-silicon pacing stays consistent with (1) backlog conversion and (2) the financing and lease dynamics embedded in the AI “fabric wall.”

Remaining performance obligations (backlog proxy: $164.6BRevenue expected within 12 months from remaining: ≈30%
Read article
CCL is the hidden bottleneck in 800G/1.6T AI racks: high-speed laminate supply—not board fab—sets how fast AI networking scales insight cover
Supply Chain
8 min read

CCL is the hidden bottleneck in 800G/1.6T AI racks: high-speed laminate supply—not board fab—sets how fast AI networking scales

As AI fabrics move from 800G toward 1.6T, the limiting factor shifts from PCB shops to the upstream materials stack that preserves signal integrity at ultra-high frequencies. Copper-clad laminate (CCL) makers are therefore pulling through demand earlier than board assemblers, because every extra layer and every tighter dielectric/trace-loss target consumes more of these low-loss laminates per board.

Business profile: CCL + prepreg focusOperating scale (latest snapshot available): EBIT margin: 22.7%
Read article
CXL “memory pooling” can monetize scarcity after HBM sells out—without buying more DRAM insight cover
Supply Chain
10 min read

CXL “memory pooling” can monetize scarcity after HBM sells out—without buying more DRAM

Micron’s DRAM/HBM tightness is forcing hyperscalers to redesign capacity access, not just capacity procurement. In that setup, the CXL memory-semantic layer (controllers, retimers, and pooling/switch fabric) is positioned to convert “scarce bytes” into software-like, elastic shared memory—creating a new equipment revenue pool that the market largely hasn’t mapped.

Micron TTM revenue level: $90.27BMicron TTM net income level: $50.47B
Read article

What to expect

Evidence-first notes with a visible point of view.

This section collects sharp takes on earnings, shareholder meetings, and market structure. Each new piece should make the thesis, the facts, and the implications obvious within the first few screens.

Expect direct analysis, not generic commentary.

Expect the data to be explicit and the argument to be easy to follow.

About UsPrivacy PolicyTerms of ServiceDisclaimerStock DirectoryETFsIndices13F HoldingsAnalyst TargetsFree ToolsTechnology MapJoin our Slack

Plutux is not an investment adviser. Market data and AI-generated analysis are for information and education only, not investment advice. Disclaimer

© Plutux Technology Limited 2026