Top 10 in Tech - What to know for Week ending March 13, 2026

1. SaaS METRIC OF THE WEEK: CAC Payback - the 'payback' period is the nuance of why we measure CAC. How long until we break even? Benchmark-wise, the negative trough is way longer than you think, so take a seat! New B2B customers, on average, (average!) take 2 years and 2 months to become profitable. This really highlights a deepening dependency on access to capital to fund a SaaS company's growth through these SaaS Cash Flow Trough. Here is an article that looks at ways to shorten CAC Payback, and this article takes an operator view on CAC Payback that includes (and celebrates) the value of net revenue retention.


2. PROMPTS: Something I learned last month - your prompts are portable, and this is one of the better things Lemkin has written lately. If customers can export the prompts that power an AI workflow (which a lot of us did last week), switching vendors becomes trivial - meaning "AI agent churn" is the real issue-du-jour and far higher than traditional SaaS lock-in ever allowed.


3. VENTURE: Who's funding what? We all know by now that AI is swallowing venture funding. Crunchbase data still shows the hottest startups now raise far larger rounds, far earlier - but the capital is concentrating into a smaller set of companies. The gap between AI winners and the rest of the startup market is widening fast. Fun fact, Tiger Global and SoftBank Vision Fund led the 2021 boom - now both have cut activity by more than 95% by deal count.


4. EMERGING OPPORTUNITIES: Venture capital isn't evenly distributing its bets anymore. PitchBook's 2025 analysis shows early-stage returns concentrating in a handful of sectors. Early-stage AI deal is up 40+% and AI median pre-money valuations hit $52.1M (up from $28.1M in Q1 23). SaaS still leads in exit probability, with a 72.1% successful exit rate. Defense tech is the quiet/not-so-quiet mover, with median early-stage valuations passing $100M.


5. EARLY-STAGE: Another report this week — this one from CB Insights, covering early-stage trends - flags agentic AI security as the category proven founders are racing to own: 10 of 21 tracked startups rank in the top 100 (and 18 of 21 average a 55% M&A probability).


6. DATA CENTRES: AI's infrastructure boom is colliding with physics and geopolitical BS. US data centers are facing all kinds of power/infrastructure shortages, grid constraints, and rising energy costs, and transformers now take 4 years to deliver! Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI just signed Trump's Ratepayer Protection Pledge to self-fund all energy costs. But guess what - it's non-binding. Side bar: Google plans on spending $185 billion on AI capex. ANNUALLY. FOR THE NEXT FRIGGIN DECADE! That's a lot of Gemini Subs.


7. VC SUBSIDIES: I have this running techism I tell anyone who will listen- enjoy new tech things as much as you can while the VC Subsidy exists: Uber, AirBnB, Netflix, Shopify, Dropbox, MealPal - all these things I got to enjoy at the VC's expense to allow blitz scaling to happen.  And of course, AI is running the same model. A $200/month Claude Code subscription might consume up to $5,000 in compute, according to the CEO of Cursor.


8. BRAND: 45% of online shoppers already use GenAI tools to compare products, and Alphabet now has 75M daily users in AI Mode, according to this Bain Survey. Brand is increasingly what the agent says it is, not what your website says - this means we've gotta design for both human users and machine decision-makers.


9. PRODUCTIVITY: Here are some incredible productivity data points - great lens. Small teams move faster because communication scales badly. Tomasz Tunguz runs the org chart math: a 150-person company has 11,175 potential communication channels; a 30-person (and AI-enabled) team producing equivalent output has 435. Anthropic generates ~$5M in revenue per employee; Cursor, $3.3M; traditional SaaS considers $200-300k strong.


10. CASE STUDY: Uber is quietly running one of the largest internal AI engineering experiments. 92% of developers use agents monthly, 31% of code is AI-authored, and 11% of PRs are opened by agents. Downsides too: AI-related costs are up 6x since 2024, and token optimization is now a dedicated priority.


POD OF THE WEEK: Design as we know it is dead? Jenny Wen (head of design at Claude) reckons so. ​