Posts

Which is the best business in AI at the moment?

I analyzed Q4 revenue data from publicly traded companies across multiple sectors—software companies, consulting firms, and hardware manufacturers to determine which segment dominates the AI market.

NVIDIA’s data center business dominated the field, generating $31b in Q4 revenue with impressive margins exceeding 70%.

In second place, Microsoft’s AI business, including Azure, is at a $3.25b Q4 revenue.

IBM reported $2b, which was a big surprise to me. Clearly, IBM’s role as the third largest generator of AI revenue is a combination of consulting and software revenue. In their most recent quarterly earnings, the company clarified the breakdown in their book of business : 20% software & 80% consulting.

During a recent Theory Office Hours with Kady Srinivasan CMO at Lightspeed Commerce, Dropbox, and Klaviyo, we discussed several powerful insights emerged on how early-stage companies should approach marketing.

Here are the highlights :

1. Define Your ICP Before Anything Else

   The most fundamental decision for any startup is determining your Ideal Customer Profile (ICP). Without clarity on exactly who you’re targeting, GTM efforts become diluted and ineffective.

   The earlier the company, the narrower it should be. With scale & capital, it can broaden.

Most startups play defense when discussing pricing with customers. They dance between asking for too little, leaving money on the table, and asking for too much, only to lose the customer’s interest. The very best companies lead their customers in that dance. They use pricing as an offensive tool to reinforce their product’s value and underscore the company’s core marketing message.

For many founding teams, pricing is one of the most difficult and complex decisions for the business. Startups operate in newer markets where pricing standards haven’t been set. In addition, these new markets evolve very quickly, and consequently, so must pricing. But throughout this turmoil, startups must adopt a process to craft a good pricing strategy, and re-evaluate prices periodically, at least once per year.

Algorithms needed for unpredictable journey. Significant compute costs, endless data processing, long periods of unexplained failures. Safe convergence doubtful. Honor & recognition in event of success.

What Shackleton might have written in 2025.

AI isn’t perfect but its productivity gains are undeniable. ServiceNow, Microsoft, and Amazon, plus nimble startups generating tens of millions in ARR with teams small enough to feed with a few pizzas all have benefitted.

The new boast of 2025? ARR per employee measured in the millions up 5x or more from the previous generation of software companies.¹ Fueled by AI.

A decade ago, most startup pitches ended with a calculation justifying the amount they sought to raise.

In other words, the raise was an output of the financial model.

But for the most sought after companies, the raise amount is disjointed from the capital needs of the business - instead it’s driven by the fundraising auction.

Great fundraisers are the teams that build the most auction pressure.

This auction dynamic, combined with venture capital’s explosive growth over the last decade, has transformed fundraising strategy. The goal is no longer to raise what you need, but to optimize for the lowest cost of capital.

October 2024 marked a critical inflection point in AI development. Hidden in the performance data, a subtle elbow emerged - a mathematical harbinger that would prove prophetic.

What began as a minor statistical anomaly has since exploded into exponential growth.

Since then AI performance has surged attaining a new trajectory, a new slope - no longer linear but geometric.

Segmenting out the models by size and type, the source of the innovation is clear. No longer model size which drove the initial wave of improvements, nor the improvements in the smaller models of the early fall.

Theory’s name isn’t just a name - it’s our ethos. We develop & test theories about the future of technology, business, & venture capital through constant experimentation.

We’re seeking an architect of the future, someone who thrives at the intersection of AI & process design.

Ideally, this person :

• enjoys architecting processes and implementing them with a team
• is relentlessly curious about new technologies & techniques
• has a good understanding or desire to learn about great sales teams
• thrives in a warm, collaborative, & urgent culture

Apply here if you’re interested.

For pre-seed to Series B founders, navigating GTM strategy, marketing, and positioning can be challenging. When should you hire your first marketing leader? How do you scale GTM efficiently? When is the right time to invest in branding?

To help founders tackle these challenges, Theory Ventures is hosting exclusive 1:1 Office Hours with Kady Srinivasan, former CMO at Lightspeed Commerce and GTM leader at Dropbox and Klaviyo. Kady has led GTM teams that scaled revenue from $150M to $600M ARR, drove successful IPOs, and optimized marketing efficiency at scale.

Cloudflare’s earnings last week revealed something more significant than just company optimism: a fundamental shift in software buyer confidence.

“However, as the quarter progressed, we saw encouraging signs that confidence is beginning to return, particularly in the U.S. Security, AI, modernization and efficiency form the word cloud we hear most often in these conversations. These themes play directly to Cloudflare’s strength.”

Sales cycles are accelerating.

“Beyond the qualitative, we saw measurable improvements in Q4. We saw a notable uptick in close rates. We saw an improvement in sales cycles. The majority of large customers I mentioned last quarter, those deals had slipped from Q3, reengaged and signed significant contracts in Q4. We crossed 3 million active developers on our platform, including Cloudflare Workers and Workers AI.”

What is the impact of AI across different levels of seniority?

Over the weekend, I read Sergey Tselovalnikov’s post on AI Impact Curves.

The software engineering curve reveals an intriguing pattern.

Junior engineers experience both benefits and risks from AI. Staff+ engineers also gain tremendous leverage at the senior end of the curve.

However, mid-level engineers see more modest impacts, primarily because they are proficient with their codebases and can write effective code adeptly.