AI’s $1 Trillion Reality Check: Earnings Put Gen AI to the Test
Welcome to Sequencr AI’s newsletter – Prompt & Circumstance – our weekly deep dive into the evolving world of Generative AI and what it means for marketing and communications.
While the world reacts to AI news in real time, we take a different approach – cutting through the noise to deliver a second-thought analysis that actually matters. Instead of knee-jerk hot takes, we break down what’s happening and what it means for you.
Let’s get into it.
TL;RD below!
AI’s $1 Trillion Reality Check: Earnings Put Gen AI to the Test
One week after more than $1 trillion in tech valuations vanished, earnings season arrived with some much-needed answers. Major AI and tech players reported their latest results, offering a clearer picture of where AI investments are actually paying off – and where they’re falling short.
This comes amid rising skepticism, fueled by DeepSeek’s sudden emergence, that AI’s economic fundamentals aren’t keeping pace with the massive investment required to sustain it. While adoption is accelerating, questions remain about whether these AI bets will translate into long-term profitability.
So, what did earnings tell us about the state of AI? The answer: It’s complicated.
Meta’s AI Strategy Delivers: Big Profits, Bigger Conviction
While Nvidia’s stock took a beating after DeepSeek's R1 model release, Meta walked away as one of the biggest winners – looking more like the early Bitcoin investor who saw the future while others hesitated.
Why?
DeepSeek’s R1 model proved that high-performing AI can be built for far less than what Big Tech is spending. That raised fears that AI models could become commodities rather than moats, hitting closed-source players the hardest – especially companies like Google, which rely on AI being difficult to replicate.
Open-source flips this equation. Instead of trying to gatekeep AI, companies like Meta ensure their models become foundational infrastructure – widely used and deeply embedded into enterprise applications. This creates network effects, where more adoption leads to faster iteration, lower costs, and stronger integration into business ecosystems.
Strong Earnings Reinforce Optimism for Meta’s Role in AI
Meta’s earnings, released on January 29, crushed any concerns about its ability to turn AI into real revenue.
In Q4 2024, Meta’s revenue surged 21% year-over-year, blowing past expectations. The driving force? Meta’s AI models optimized ad targeting, leading to a 6% increase in ad impressions and a 14% rise in average ad prices.
Unlike companies still struggling to monetize AI, Meta has already built an AI-driven ad engine that’s delivering tangible ROI.
For brands and advertisers, this underscores why Meta’s AI-driven ad targeting is a must-watch. AI-powered personalization is becoming table stakes, and Meta’s ability to improve ad efficiency while increasing prices suggests that brands will need to get smarter about leveraging AI in their campaigns.
Winners and Losers in the Gen AI Race Becoming Clearer?
But Meta wasn’t the only company that made a statement last week. While some players proved their AI strategy is working, others struggled to justify their massive investments.
Winner – Microsoft
Microsoft appears to be the frontrunner in the enterprise race for Generative AI adoption. The old maxim, "You don’t get fired for hiring IBM," seems to apply here –except this time, it’s Microsoft that’s benefiting. The company reported a 175% increase in its AI business revenue year-over-year last week. Their relentless focus on security, compliance, and enterprise-grade AI has created just enough uncertainty and doubt around competitors to solidify their growth, even as CoPilot’s functionality lags behind options like ChatGPT.
Winner – Salesforce Delivers on Benioff’s Bluster
Marc Benioff hasn’t been pulling punches. At Dreamforce 2024, he ripped into Microsoft’s Copilot, calling it “basically the new Microsoft Clippy” –a brutal throwback to one of tech’s most infamous flops. Last week, at Davos, he doubled down, saying “Everybody knows that Microsoft has disappointed the world with Copilot, and I mean, that's just a repackaged ChatGPT. You know that.”
His remarks didn’t go unnoticed. Microsoft’s comms offered a spirited defense with some interesting commentary on LinkedIn.
Salesforce has been pushing hard into AI agents, embedding its Einstein product across its product suite and launching Agentforce. Benioff has argued that AI should do more than just assist – it should actually get work done.
Loser – Amazon’s Big Spend and Capacity Problems
Amazon’s stock dropped last week as investors worried about how much it’s spending on AI and whether it can keep up with demand. The company is pouring $100 billion into AI-related infrastructure in 2025.
AWS has long been the backbone of cloud computing, but there’s a problem – the company is running into hardware shortages and power supply issues, making it more challenging to scale as fast as businesses need.
In an industry where speed matters, delays could push companies to look for alternatives. The silver lining for the likes of Amazon is that there is more demand than capacity.
Loser – Is Google Losing Steam?
Investors were expecting a better AI story from Google. The company’s AI-driven cloud revenue underperformed, leading to a 7.2% drop in the stock last week.
But perhaps the bigger red flag? Google launched new AI models and barely anyone noticed.
The company introduced Gemini 2.0 Flash, a faster, more cost-efficient model meant to compete with OpenAI and Meta, as well as an upgraded Gemini 2.0 Pro. These should have been major announcements, yet the market barely reacted.
For a company that has long been seen as an AI pioneer, that’s a problem. It is hard to pinpoint what the issues is. In tests, Gemini’s models perform just as well as industry leaders, but people who use AI regularly say they often miss the mark, don’t follow instruction well, or deliver incomplete results compared to ChatGPT or Claude.
The $6M Myth Unravels?
Last week’s earnings and Nvidia’s stock rebound raise an important question - was the DeepSeek market freakout overblown?
It’s starting to look that way. DeepSeek’s claim that it trained a cutting-edge AI model for just $6 million sounded like a game-changer – one that could rewrite the economics of AI development. But as the dust settles, it’s looking more like a headline-grabbing half-truth.
Analysts now estimate that the company spent closer to $1.6 billion on hardware alone, acquiring as many as 50,000 high-end Nvidia Hopper GPUs. And that’s before accounting for data processing, research, and infrastructure costs, which could push the real price tag even higher.
So where did the $6 million figure come from? It appears to only reflect the GPU time needed for a single pre-training run, leaving out all the other costs that go into making a model commercially viable.
And then there’s the question of how DeepSeek got its hands on those GPUs in the first place. With U.S. export controls restricting the sale of high-end Nvidia chips to Chinese firms, reports suggest that DeepSeek may have sourced its GPUs through intermediaries in Singapore, potentially skirting trade restrictions.
DeepSeek’s approach to training AI is genuinely innovative and could shape the future of model development. For marketing and communications leaders, DeepSeek’s rise signals a shift in AI - no single model will dominate, and businesses now have real choices. The era of being locked into one AI provider – whether OpenAI, Google, or Meta – is over before it even started. The AI landscape is diversifying, and that’s a good thing.
DeepSeek Raised Doubts About AI Valuations – But OpenAI Remains Undaunted
When DeepSeek burst onto the scene, it called into question the sky-high valuations of AI startups. Critics argued that OpenAI, Anthropic, and others may not be worth nearly as much as investors think.
If a relatively unknown startup can build a powerful, cost-efficient AI model in secret, what’s stopping others from doing the same?
But OpenAI isn’t backing down. Less than three months after raising $6.6 billion at a $157 billion valuation, the company is now reportedly looking to raise up to $40 billion – at a staggering $300 billion valuation. This would make OpenAI the second most valuable startup in the world, trailing only SpaceX.
This raises the $300 billion question: Is OpenAI still the clear leader, or just the most expensive bet in an AI landscape that’s getting more competitive by the day?
That’s all for this week, folks.
We’ll be back next week with more AI insights, second-thought analysis, and a breakdown of what actually matters in the Gen AI race.
TLDR
Meta came out strong post-DeepSeek panic, proving that open-source AI models can be a winning strategy. AI-driven ad targeting is working, driving revenue growth.
Microsoft and Salesforce are curent AI winners. Microsoft dominates enterprise AI adoption despite CoPilot’s weaker functionality, while Salesforce is pushing AI agents that actually "get work done."
Amazon and Google struggle. Amazon faces infrastructure bottlenecks despite high demand, and Google’s Gemini 2.0 launch barely registered with users or investors.
DeepSeek’s claim that it trained a model for $6M was misleading, but its training innovations could reshape AI development.
OpenAI isn’t backing down. Despite concerns over AI valuations, it’s reportedly raising $40 billion at a $300 billion valuation - can these valuations be sustained?