Newsletter comparison
Citrini Research vs FUNDA
Short answer: Choose Citrini Research for scenario design and thematic risk. Choose FUNDA for company-level AI-infrastructure diligence. Read both when a big thesis needs to survive contact with capex plans, suppliers, and earnings.[1][2][3][4][5][6]
An independent guide. Scottie isn't affiliated with, endorsed by, or sponsored by Citrini Research and FUNDA. Each publication owns its name, writing, and subscription terms.
A brief from both newsletters · July 31, 2026
See both newsletters in one brief
See what made the brief, what didn't, and the original links behind every included story.
In this brief: Scottie kept the selected stories separate because they added different value.
Reader priorities
An investor wants company and supply-chain evidence for the long-run capital, infrastructure, and competitive effects of the AI buildout.
These are illustrative priorities, not a customer’s data.
What made the brief
- Citrini Research1 read · 1 included
- FUNDA19 read · 9 included
Scottie
Scottie example brief
July 31, 2026 · Executive brief
01 / The rundown
- xAI locked in 8GW of hardware demand through 2027, highlighting severe power and permitting bottlenecks across the data center supply chain.
- TSMC upgraded full-year revenue growth above 40% while accelerating its 2nm process ramp to satisfy relentless chip demand.
- ASML raised its 2026 revenue guidance to 45 billion euros, confirming sustained spending on foundational lithography capacity for next-generation hardware.
02 / The briefing
01 / main
xAI Locks In Massive 8GW Hardware Pipeline Through 2027
The takeaway: xAI is securing an 8GW footprint via 13k GB300 and 15k Rubin racks. However, supplier capacity currently underwrites only 3.5 to 5GW, forcing reliance on a $1.465B Chinese turbine deal to overcome power generation gates.
Concrete details
- Colossus 2 reaches 946MW, making it the largest single-site AI data center globally.
- Channel checks reveal a supply gap, with vendors underwriting 3.5–5GW against xAI's 8GW demand.
Why it matters for this reader: Gives you clear hardware demand visibility to evaluate whether power supply bottlenecks will delay long-term mega-cap infrastructure deployments.
Original sourcesFUNDA
02 / main
TSMC Accelerates 2nm Ramp Up Amid AI Wafer Demand
The takeaway: TSMC raised its full-year sales growth outlook above 40% as AI demand far outstrips supply. A faster 2nm process rollout will temper second-half margins but accelerates advanced foundry capacity expansion.
Concrete details
- TSMC projects a 2026-2028 capacity compound annual growth rate above 70% for 2nm chips.
- Faster 2nm production will dilute second-half 2026 gross margins by 3% to 4%.
Why it matters for this reader: Provides primary foundry evidence that leading-edge node expansion is accelerating, validating your thesis on multi-year hardware capital intensity.
03 / main
ASML Raises Revenue Guidance on EUV Expansion
The takeaway: ASML lifted its 2026 revenue target to EUR 43bn–45bn as fab upgrades boost capacity. Higher lithography tool shipments reflect persistent advanced semiconductor investments despite market volatility.
Concrete details
- ASML projects EUV shipments of 66 units in 2026, 95 in 2027, and 109 in 2028.
- Installed Base Management business revenue is expected to grow over 30% in 2026.
Why it matters for this reader: Confirms foundational equipment spending remains resilient, offering you concrete proof of durable long-term capital allocation at the lithography bottleneck.
04 / main
Kimi K3 Architecture Shifts Storage and Memory Needs
The takeaway: Kimi K3 uses NAND offloading for 75% KV cache compression, but its 2.8-trillion parameter size still requires high-bandwidth supernodes. The efficiency gain actually enables broader model scaling without trimming overall DRAM or HBM demand.
Concrete details
- Kimi K3 features 2.8 trillion total parameters, activating 16 of 896 experts per token.
- Deployment guidelines specify high-bandwidth supernodes containing at least 64 hardware accelerators.
Why it matters for this reader: Helps you assess how evolving algorithm choices impact long-term memory demand rather than assuming software optimizations crush hardware sales.
05 / main
AI Market Volatility Masks Hidden Supply Chain Value
The takeaway: Recent market selloffs reflect unwinding leverage rather than shrinking compute spending. Aggregate token spend continues to grow, creating potential investment opportunities among overlooked hardware designers, toolmakers, and supply chain partners.
Concrete details
- Aggregate token spend continues rising rapidly despite recent momentum-driven equity market selloffs.
- Kimi K3 open-source model remains compute-constrained, proving algorithmic gains do not eliminate infrastructure needs.
Why it matters for this reader: Directs your portfolio analysis toward underappreciated supply chain links as market momentum shifts away from crowded consensus trades.
Original sourcesCitrini Research
06 / main
Microsoft Copilot Seats Surge to 37 Million
The takeaway: Microsoft added 15 million paid M365 Copilot seats last quarter, driven by enterprise expansion. To support this growth, the cloud giant is expanding data center capacity and GPU leasing while restructuring its product organization.
Concrete details
- Paid M365 Copilot seats jumped from 22 million in March 2026 to 37 million by late June.
- User keep rates for complex tasks improved from roughly 40% to 70% quarter-over-quarter.
Why it matters for this reader: Connects software monetization progress with ongoing cloud infrastructure spending, helping you verify that end-user demand supports continued capital deployment.
Action items
- Audit semiconductor holdings to account for ASML's upgraded 2026 EUV shipment forecasts and expanding tool backlog.
- Model power supply exposure across chip manufacturers following xAI's turbine orders to address data center capacity shortages.
- Re-evaluate memory stock valuations to separate architectural KV cache efficiency gains from baseline HBM hardware demand.
See every issue behind this brief
xAI 8GW Buildout Mapping and Hardware Commitment
- Deep|SPCX: xAI 8GW Buildout MappingFUNDA · Included
ASML EUV Capacity Expansion and Shipment Guidance Boost
TSMC CapEx Upside Surprise and AI Manufacturing Capacity
- Review|TSM 2Q26: CapEx Upside Surprise; 3Q26 Gross Margin Below ExpectationsFUNDA · Included
- Preview|TSMC 26Q2: Expect Another Beat-and-RaiseFUNDA · Included
ATI Starship superalloy demand
Meta 2027 capex plans
Meta AI compute commercialization
- Weekly|Meta NeoCloud Fears Overdone, Key AI Narrative Shifts, GPU Spot Price Noise, SPCX xAI EconomicsFUNDA · Read, not included
Microsoft Enterprise AI Monetization and Data Center Expansion
Memory Infrastructure Implications of Kimi K3 KV Cache Efficiency
Value Accrual Dynamics Across the AI CapEx Chain
- All Along the AI WatchtowerCitrini Research · Included
Kimi K3 launch impact
Kimi K3 performance review
- Research|LLM: Kimi K3 - Scaling Still Works; An Expensive Model Competing at Front TierFUNDA · Read, not included
xAI Grok 4.5 release
Meta NeoCloud business model
- Research|META: Building a NeoCloud and Continuing to Rent Capacity are not ContradictoryFUNDA · Read, not included
Palantir 2Q26 growth momentum
- Preview|PLTR: 2Q26 Growth Momentum Remains StrongFUNDA · Read, not included
ServiceNow 2Q26 channel feedback
- Preview|NOW 2Q26: New Pricing Pull-Forward Tailwind Drives the Beat, AI Still LukewarmFUNDA · Read, not included
ServiceNow 2Q26 earnings beat
- Review|NOW 2Q26: Pull-Forward Expected, Restrained Full-Year GuideFUNDA · Read, not included
The useful difference
The difference that matters
Citrini Research develops broad future scenarios and thematic risks. FUNDA works closer to individual companies, infrastructure projects, products, and the supply chain behind reported growth.[1][2][3][4][5][6]
What each is best for
When it’s worth reading both
The combination is useful when a persuasive 2028 story must be checked against what companies are building and reporting now. Scottie can link the horizon without disguising the gap in certainty.[1][2][3][4][5][6]