OpenAI

Anduril

OpenAI

OpenAI makes money primarily through API access and subscriptions. The API charges developers per token (roughly per word) for using GPT models in their applications.

ChatGPT Plus costs $20/month for individual users, ChatGPT Team is $25-30/user/month, and ChatGPT Enterprise is custom-priced. Microsoft pays OpenAI licensing fees and also resells OpenAI models through Azure OpenAI Service.

OpenAI's annualised revenue ran past $13 billion during 2025 and kept climbing from there, which is the kind of growth curve that normally only exists in a pitch deck.

Anduril

Anduril flips the traditional defense business model. Instead of cost-plus contracts where the government pays for development, Anduril invests its own capital in R&D and sells finished products.

This means they own the intellectual property and can sell the same platform to multiple customers — the US military, allied nations, and potentially commercial clients.

Revenue comes from hardware sales (drones, autonomous vehicles, sensor towers), software licensing (the Lattice operating system), and service contracts for deployment and maintenance. The company has won contracts with the Department of Defense, Department of Homeland Security, US Special Operations Command, and allied militaries including the UK and Australia.

The venture-funded approach lets them move at startup speed. While Lockheed might take 7 years to develop a new system, Anduril can prototype in months and iterate based on field feedback.

The trade-off is massive upfront investment — they have raised more than $11 billion in venture capital to fund this approach.

OpenAI

OpenAI was founded in December 2015 as a nonprofit AI research lab. The founding donors — including Elon Musk, Sam Altman, Peter Thiel, Reid Hoffman, and Jessica Livingston — pledged $1 billion with a mission to build artificial general intelligence (AGI) that would benefit all of humanity.

The idea was that AI was too important and too dangerous to leave in the hands of Google alone.

Sam Altman became chairman while Greg Brockman (former CTO of Stripe) became president. Ilya Sutskever, one of the most respected AI researchers alive, left Google Brain to become chief scientist.

The early team was stacked with world-class researchers who published their work openly — hence "Open" AI.

But AI research turned out to be staggeringly expensive. Training large models required millions of dollars in compute.

In 2019, OpenAI created a "capped-profit" subsidiary — investors could earn up to 100x their money, but profits beyond that would flow to the nonprofit. Microsoft invested $1 billion.

The mission was still to save humanity. The method now involved making a lot of money first.

Anduril

Palmer Luckey was already famous — and controversial — before Anduril. He founded Oculus VR at 18, sold it to Facebook for $2 billion at 21, and then was fired in 2017 after a political donation scandal.

He was 24 years old, already worth hundreds of millions, and suddenly had nothing to do.

Luckey teamed up with Trae Stephens, a Founders Fund partner who had previously worked at Palantir and as a member of the Trump transition team's Department of Defense group. They saw the same problem from different angles: the US military was spending billions on outdated technology from legacy contractors (Lockheed, Raytheon, Boeing) while China was rapidly modernizing.

The Pentagon's procurement process was broken — it took 10-15 years and billions of dollars to develop and deploy new weapons systems.

Anduril was founded in mid-2017 with a radical approach: build the technology first with venture capital, then sell finished products to the government. Traditional defense contractors get cost-plus contracts — they bill the government for development costs plus a margin, which incentivizes slow development and cost overruns.

Anduril said: we'll fund our own R&D, build the product, and sell it off the shelf. If it doesn't work, we eat the cost.

The name comes from the reforged sword of Aragorn in Tolkien's Lord of the Rings — a weapon that was broken and made new.

OpenAI

ChatGPT's launch in November 2022 was the growth strategy — it just wasn't planned that way. The team expected a modest research preview.

Instead, ChatGPT hit 1 million users in 5 days and 100 million monthly active users in 2 months, making it the fastest-growing consumer application in history. The product went viral because it felt like magic — for the first time, anyone could have a natural conversation with a machine that seemed to understand them.

The Microsoft partnership provided distribution at massive scale. Microsoft integrated OpenAI models into Bing, Office 365 (Copilot), GitHub (Copilot), and Azure.

Overnight, hundreds of millions of Microsoft users had access to OpenAI technology. Microsoft's $13 billion investment was the largest AI bet in history and gave OpenAI nearly unlimited compute.

The API created an ecosystem. Thousands of startups built products on top of OpenAI's models — from customer service bots to coding assistants to content generators.

Each API customer locked themselves into OpenAI's ecosystem, creating switching costs and recurring revenue.

Anduril

Anduril's growth strategy is classic disruption — enter at the low end of the market with cheaper, faster products and expand upward. They started with border security (relatively low-stakes) and moved into counter-drone systems (active combat relevant), then into autonomous vehicles and munitions (core defense).

International sales are a major growth vector. Anduril has contracts with the UK Ministry of Defence, the Australian Defence Force, and other Five Eyes allies.

The AUKUS defense pact between the US, UK, and Australia specifically calls for technology sharing in areas where Anduril specializes.

The Ukraine war was an inflection point. It demonstrated that small, cheap, autonomous drones could be decisive in modern warfare — exactly the kind of systems Anduril builds.

Suddenly, every military in the world wanted what Anduril was selling, and wanted it fast. The order book has grown into the billions and the company has raised at valuations no defence startup had ever reached.

OpenAI

The board crisis of November 2023 nearly destroyed the company. The nonprofit board fired Sam Altman as CEO on a Friday, citing a loss of confidence.

Within 48 hours, 95% of employees threatened to quit and follow Altman to Microsoft. By Tuesday, Altman was reinstated and the board was restructured.

The incident exposed the fundamental tension between OpenAI's nonprofit governance and its for-profit ambitions — a tension it finally addressed in late 2025 by restructuring the business as a public benefit corporation sitting under the original nonprofit.

The cost of training frontier models is eye-watering. Each new GPT generation costs hundreds of millions to train.

OpenAI is reportedly spending over $7 billion annually on compute. The company is burning through cash faster than almost any startup in history, which is why it keeps raising at higher and higher valuations.

If revenue growth slows before costs stabilize, the math gets ugly.

Safety concerns are not going away. Multiple prominent researchers have left OpenAI over disagreements about the pace of development versus safety research.

Ilya Sutskever, the chief scientist who was central to the board's decision to fire Altman, left in 2024 to start a safety-focused AI lab. The public debate about whether OpenAI is moving too fast — and whether its safety commitments are genuine — grows louder with every capability improvement.

Anduril

The ethical debate around autonomous weapons is constant and unavoidable. Anduril builds systems that can identify and engage targets with varying degrees of human oversight.

Critics argue this is a step toward fully autonomous killing machines. Anduril maintains that a human is always "in the loop" for lethal decisions, but the line between "in the loop" and "on the loop" (supervising but not directly controlling) is blurry.

Recruiting is both an advantage and a challenge. Anduril pays Silicon Valley salaries and offers startup equity, which attracts top engineers who might never consider working for Raytheon.

But some engineers refuse to work on weapons systems on principle. Google famously dropped Project Maven (a Pentagon AI contract) after employee protests.

Anduril leans into the controversy — they explicitly look for people who are comfortable building defense technology.

Scaling manufacturing is the next hurdle. Software companies scale effortlessly.

Hardware companies that build drones, missiles, and autonomous vehicles need factories, supply chains, and quality control at defense-grade standards. Anduril is building a massive manufacturing facility to produce thousands of autonomous systems, but transitioning from prototype to mass production is where many defense startups fail.

OpenAI

ChatGPT is the consumer chatbot — the product that made AI mainstream overnight. The flagship GPT models are multimodal, handling text, images, and audio in one system.

The OpenAI API lets developers integrate GPT into any application. DALL-E generates images from text descriptions.

Whisper transcribes and translates audio. Sora generates videos from text prompts.

GPT Store lets users create and share custom GPT agents. ChatGPT Enterprise gives businesses a private, secure version of ChatGPT with admin controls and no data training.

Anduril

Lattice — an AI-powered operating system that fuses sensor data from multiple sources into a single real-time picture of the battlefield. Think of it as the central nervous system that connects all of Anduril's hardware.

Ghost — a family of small autonomous aircraft (drones) designed for surveillance, electronic warfare, and strike missions. Ranging from handheld to medium-altitude.

Anvil — an autonomous counter-drone system that physically intercepts enemy drones by ramming into them mid-air. Yes, a kamikaze drone that kills other drones.

Sentry Tower — autonomous surveillance towers originally deployed on the US-Mexico border for border security, using AI to detect and classify objects and people. Altius — a family of tube-launched autonomous munitions that can loiter over an area and strike targets with precision.

OpenAI

Microsoft ($13B), Thrive Capital, Khosla Ventures, Sequoia Capital, Founders Fund, Tiger Global, SoftBank, a16z

Anduril

Andreessen Horowitz has been the most prominent backer, leading multiple rounds. Founders Fund (Peter Thiel's firm) invested early — Trae Stephens was a Founders Fund partner before co-founding Anduril.

General Catalyst, Valor Equity Partners, and 8VC (Joe Lonsdale, another Palantir co-founder) also invested. The Series F in 2024 valued the company at $14 billion, and successive rounds took it to $30.5 billion in 2025 and past $60 billion after that.

Fidelity, T. Rowe Price, and Sands Capital participated in later rounds.