Welcome to Whale Watching - the monthly column by
Franz Schrepf where we dive into gigantic alliances to learn what it takes to make a splash in your industry. In this blog we’ll discuss the amazingly awkward partnership between Microsoft and OpenAI.

The launch of ChatGPT rocked the world. It sparked a new space race for AI and catapulted OpenAI into the global news. It’s the fastest growing consumer product in history, with
+100M sign ups in less than two months. So why would this rising star sell a 49% stake to Microsoft for a “meager” $10B investment? The Wall Street Journal called it “the awkward partnership leading the AI boom” that may wipe out the human race. The stakes are high, so let’s dive deeper.

The Bad: Open AI & Microsoft’s trouble

Underneath the hype, OpenAI was in deep trouble. Yes, ChatGPT was cutting-edge tech. But two huge challenges could have killed it anytime:

• Resources as a moat:
  AI models have no moat. Advancements happen so fast, ChatGPT will be outdated in no time. The only way to stay on top is to train your models on ever larger amounts of data. Training is expensive, and this is where OpenAI got in trouble: it made a lot of noise. The former AI incumbent, Google, took notice. It could use the full might of its Google Cloud and Deepmind teams to catch up and outpace OpenAI.

• Distribution: OpenAI had no distribution. As a startup, it relied on hype and virality to spread the word. But when the hype fades, then what? ChatGPT was the perfect feature, but not a product. It had the potential to kill Google search, but not if Google puts its own tool in front of billions of users first.
  

Similar to our previous post about
the $100M+ Facebook and Microsoft alliance, Google was the enemy. And again, Microsoft struggled with its own issues:

• Bing:
  Have you ever used Bing? Like… intentionally? With a
  market share of 9%, most people haven’t. Google has dominated search for close to 20 years. The lines were drawn. Microsoft needed to disrupt the industry to deal Google a critical blow.

• Clippy and friends: Remember Clippy, the Word helper? Or Cortana, Microsoft’s version of Siri? No? How about
  Tay, the Twitter bot who had to be shut down because it started to spew antisemitic slurs? No major player had missed the boat so many times in AI.

LESSON: The enemy of my enemy is my friend. Mutual competitors can provide the incentives and urgency to form strategic alliances.

The Good: distribution, finance, and the cloud.

On Jan 23rd 2023,
both companies announced a deal that looked like a match made in heaven. Or a pact with the devil.

Microsoft was a strategic investor. Few companies (Google, Amazon, Microsoft, Oracle) have data centers at the scale OpenAI needed to train their models. Technically, OpenAI could have taken money from any investor and simply paid one of the big four. But in exchange for exclusivity, OpenAI got favorable rates for Microsoft Azure’s cloud hosting. No outside investor could provide this level of cost efficiency.

The “killer feature” of the alliance was Microsoft’s distribution. OpenAI knew that ChatGPT is a feature, not a destination in itself. When added into all of Microsoft’s products, it can meet users where they already are and unleash its full potential across endless real-world use cases.

Lesson: Money is a commodity. Distribution and product drive strategic alliances.

What a deal! Microsoft won a huge customer and access to cutting-edge AI. And now OpenAI can give Google a run for its money.
The failed announcement of Google’s own chat bot, Bard, was the cherry on top. Within weeks, $100B+ was wiped off of Google’s market cap!

Would you spend $10B to tank your main competitor’s stock by $100B and send them into a frenzy? Sounds good to me. But this partnership wasn’t built overnight. The groundwork was laid a long time ago.

Between 2019 and early 2023,
Microsoft had already invested $2B+ into OpenAI. Roughly half of the investment was in the form of Azure cloud computing credits. The teams also worked on several other projects, like an Azure-hosted, Open-AI-co-designed
super computer launched in 2021.

The turning point in the relationships, however, was detailed in
Elon Musk’s biography when Sam Altman, OpenAI’s CEO, approached Bill Gates with an early version of ChatGPT. Unimpressed, Gates told them to come back when they can complete a biology exam. He thought this would take several years to achieve. Instead, the team came back three months later. That’s when Microsoft went all in.

LESSON: Trust is built through consistency over time. When the stakes are sky-high, you need a solid track record.

The Ugly: Deal Structure, core competencies, oversharing

If a partnership sounds too good to be true, it likely is. This is where the OpenAI and Microsoft alliance gets weird.

Deal Structure

OpenAI was originally founded as a non-profit organization with the mission to “ensure the creation and adoption of safe and beneficial AGI (artificial general intelligence).”

When it became clear that more money was needed to achieve this goal, OpenAI decided to create an additional for-profit company. The result is a deal structure which
The Information called “bonkers”. It looks like this:

If you’re confused, here’s the summary: Microsoft, other investors, and employees will receive most of the profits of OpenAI until a cap is reached. Then, all profits will go to the OpenAI nonprofit.

You might be wondering, “What’s the cap?”. It’s a 100x return on investment! When the news broke, OpenAI cofounder Greg Brockman tried to defend the deal on Reddit which lead to this hilarious exchange:

Greg: “We’ve negotiated a cap with our first-round investors that feels commensurate with what they could make investing in a pretty successful startup (but less than what they’d get investing in the most successful startups of all time!).”

User: “I don’t know if the best response to ‘we’re not happy that it’s being structured as a for-profit company’ is ‘yea but we could’ve made even more money!’....”

I mean, I get it. Training AI is expensive, and the people who foot the bill want to get paid if it all works out. At the same time, AGI has the potential to destroy all jobs. It could be the most valuable asset of all time. 100x seems like a lot, but if the potential upside is unlimited, you’ll want to have a cap. It’s just awkward.

Lesson: Strategic alliances require creative solutions to unique challenges.

Oversharing

Another interesting aspect is that Microsoft will “only” own 49% of OpenAI in exchange for a $13B investment. Under normal circumstances, giants like Microsoft would swallow the entire startup. But with all the spotlight on AI, an acquisition would be guaranteed
to trigger an antitrust lawsuit.

This opens up a whole new issue: who owns OpenAi’s source code, its most valuable resource? You may think the answer is OpenAI, but
Elon Musk disagrees on Twitter:

“ChatGPT is entirely housed within Microsoft Azure. When push comes to shove, they have everything, including the model weights.”

The Wall Street Journal reported that people within Microsoft have complained that OpenAI doesn’t allow most Microsoft employees to access their source code. Most! Which means some employees DO have access. Again, this is awkward, especially if you’re dealing with a company like Microsoft which hasn’t hesitated to crush smaller competitors in the past. Remember Netscape?

Lesson: Strategic partners need to protect their IP and cornered resources.

Core Competencies

Here’s the real issue with this deal: AI will be a huge part of every software product. Soon it will be the main reason people will buy products from Microsoft and OpenAI. It’s a core competency for both companies. As Oren Etzioni, board member and former CEO of the nonprofit research organization the Allen Institute for Artificial Intelligence, puts it:

“What puts them in more of a collision course is both sides need to make money. The conflict is they’ll both be trying to make money with similar products.” (Source)

This is already happening.
Microsoft and OpenAI sales teams sometimes pitch the same customers. The Bing team was also upset when OpenAI launched ChatGPT several weeks before the Bing Chatbot was ready. OpenAI stole their thunder.

What makes this mix especially explosive is Microsoft’s reduced investment in its own AI initiatives and increased reliance on OpenAI for the development of new AI tools. Microsoft’s tools can do exactly the same as OpenAI’s tools.

When companies with similar products partner, conflict is inevitable. But for now, this overlap and awkwardness won’t matter too much. Regardless of which sales team wins a customer, the profit goes to Microsoft. But if there’s ever a rift, MS has too much riding on OpenAI to let it slip.

Lesson: Where possible, partners should never outsource their core competencies.

My assumption is that Microsoft had already given up hope on its in-house AI efforts. OpenAI was the last chance to jump onto the AI bandwagon that may reshape all industries. When you need to add capabilities fast, partnerships are the way to go.

Even if they’re awkward.

Franz Schrepf is the Head of Strategic Partnerships at
StreamYard. He’s also the host of The Partner Ship, a bi-weekly LinkedIn Live Show where he interviews the world’s leading partnerships experts.