Deep Signals
Flagship long-form analysis of technology, markets, and strategy.
The Great Inference Pivot: Why Every Cloud Provider Is Rebuilding Its AI Stack
Training dominated the first wave of AI infrastructure investment. Now the real war — and the real money — is in serving models at scale.
Power Hungry: AI's Insatiable Energy Demand Is Reshaping the Global Power Grid
The collision between exponential AI compute demand and finite energy infrastructure is forcing the technology industry into the power generation business — with nuclear, natural gas, and grid politics at the center.
The Deflation Engine: AI Inference Costs Are Falling Faster Than Anyone Modeled
Quantization, speculative decoding, hardware competition, and model distillation are compounding into an inference cost decline that will reshape which AI applications are economically viable — and when.
The DX Moat: Why Developer Experience Is Becoming the Decisive Competitive Advantage in AI
As model capabilities converge, the companies winning the AI market are those that make their models easiest to build with — and the API is becoming the product.
The Sovereign AI Race: Why Every Nation Wants Its Own AI Stack
From the EU's industrial policy to Gulf state megaprojects to India's public infrastructure play, the global race for sovereign AI is redrawing the geopolitical map of technology — with consequences the industry has not yet priced in.
The Memory Wall: Why Bandwidth — Not Compute — Is the Binding Constraint in AI Hardware
The AI industry spent three years obsessing over compute. The bottleneck that actually determines inference economics is memory bandwidth, and the implications reshape the entire semiconductor landscape.
The Pilot Trap: Why Enterprise AI Projects Stall and What Separates the Companies That Scale
Most enterprise AI initiatives never graduate from pilot to production — the barriers are organizational and economic, not technical, and the companies that overcome them share a distinct playbook.
The Agent Framework Wars: Why AI's Next Platform Battle Has Already Started
LangChain, CrewAI, AutoGen, and the major labs are all racing to define how AI agents get built — and the winner will control the most consequential software abstraction layer since the web browser.
The Middleware Opportunity in AI: The Unsexy Layer Where the Money Is
Between foundation models and end-user applications lies a massive and growing market for AI middleware — guardrails, observability, evaluation, orchestration, and gateway infrastructure that enterprises actually need to deploy AI in production.
AI and the Quiet Revolution in Scientific Discovery
While the tech industry debates chatbots and agents, AI is delivering its most transformative results in science — protein structure prediction, drug discovery, materials science, and weather forecasting are being fundamentally reshaped.
The Economics of Foundation Model Companies: Revenue, Burn Rates, and the Path to Profitability
The foundation model business is structurally brutal — massive capital requirements, declining unit economics, and intense competition mean most of today's AI labs will not survive as independent companies.
Why Autonomous Coding Is Closer Than You Think
The trajectory from autocomplete to autonomous software engineering agents is accelerating — Cursor, Devin, Claude Code, and GitHub Copilot Workspace represent intermediate steps toward a fundamental shift in how software is built.
The Database Wars: How AI Is Reshaping Data Infrastructure
Vector databases, RAG architectures, and the AI application stack have ignited a new database war between purpose-built startups and entrenched incumbents — and the outcome will determine how AI applications access knowledge.
China's AI Ecosystem Is More Advanced Than Most People Think
Behind the chip export controls, a parallel AI universe is maturing rapidly — DeepSeek, Qwen, Baidu ERNIE, and dozens of capable open-weight models are reshaping assumptions about who leads in AI.
The Great AI Talent War: Why Compensation Is Reshaping the Tech Industry
The fight for AI researchers and engineers has created a two-tier compensation system in tech, with structural consequences for startups, incumbents, and the geographic distribution of innovation.
Why Cybersecurity Is the Most Underrated AI Use Case
While attention concentrates on chatbots and content generation, AI is quietly transforming cybersecurity — automating threat detection, accelerating incident response, and reshaping the arms race between attackers and defenders in ways that will define the security landscape for a decade.
AI and the Future of Search
AI-powered answer engines are dismantling the link-based search paradigm that has dominated the internet for two decades, threatening the $200 billion search advertising market and forcing Google into the most consequential strategic pivot in its history.
The Rise of Small Language Models
While frontier labs chase ever-larger models, a quiet revolution in small, efficient language models is reshaping enterprise AI deployment — proving that for most practical applications, smaller is not just cheaper but better.
Cloud Repatriation: Why Some Companies Are Leaving the Cloud
After a decade of aggressive cloud migration, a growing number of organizations are pulling workloads back to on-premises infrastructure, driven by cost overruns, data sovereignty requirements, and the realization that the cloud is not always the most economical choice.
The New Economics of AI Training Runs
As frontier AI training runs cross the $100 million threshold and climb toward $1 billion, the economics of who can afford to build foundation models are reshaping industry structure, competitive dynamics, and the geography of AI power.
Why Apple's AI Strategy Is Different From Everyone Else's
While competitors race to build the largest cloud-based AI models, Apple is making a deliberate bet on on-device inference, privacy-first architecture, and deep hardware-software integration that could define a fundamentally different AI paradigm.
The Scaling Debate: Are We Hitting Diminishing Returns?
The scaling laws that propelled AI from novelty to near-human capability are showing signs of strain, forcing the industry to reckon with whether brute-force compute increases can continue delivering proportional gains.