David Eberle

Microsoft just pulled Claude Code for 100,000 engineers because it got too expensive. The irony: they invested $5B in Anthropic, but token-based billing spiraled out of control. When the biggest AI investor admits the tools cost more than the engineers, the hype meets unit economics.

Salesforce reports earnings today. Agentforce ARR up 169% to $800M. Stock still down 51% from peak. Either AI agents expand SaaS as a digital workforce paid per resolution, or seat-based cannibalization crushes the core business. The market is voting for option 2.

Nvidia's Jensen Huang just called out tech CEOs for being "lazy" when they blame AI for layoffs. "AI has just arrived. How is it possible they're already losing jobs?" Either AI is already replacing labor, or executives are using a $3.6T hype cycle to hide standard cost-cutting.

Meta just fired 8,000 people and reassigned 7,000 to train AI. The irony: your former colleagues are training your replacement. But replacement is a trap. The companies winning the next decade won't have the smallest payrolls. They'll have the best-armed humans.

OpenAI is preparing for a $1 trillion IPO while projecting a $14 billion loss this year. They aren't building a software company. They are building an infrastructure monopoly. The era of the unprofitable software startup is dead. The era of the AI monopoly is here.

Salesforce just surveyed 3,075 enterprise leaders. 66% are now using AI customer service agents. But only 44% of consumers actually trust them. Deflection is not resolution. Are you deploying AI agents to help your customers, or just to avoid talking to them?

Google just killed the search box. 2.5 billion users are about to search differently. At I/O, they replaced the ten blue links with generative UI and autonomous agents. The SEO era is over. The internet is moving from information retrieval to action execution.

OpenAI just lost enterprise market share for the first time. ETR data shows OpenAI dropped from 62% to 56% adoption, while Anthropic's Claude more than doubled from 21% to 48%. The battle isn't over chatbots anymore. It's over coding assistants. And developers are choosing Claude.

Microsoft spent $37.5 billion on AI last quarter. But 96.7% of their users are rejecting it. A former VP just leaked that Copilot has a dismal 3.3% adoption rate. The lesson: You cannot force product-market fit through distribution. Build what customers actually want.

Figma was supposed to be the ultimate SaaS success story. Now they are facing an existential crisis. When AI can generate an entire UI from a text prompt, the value of a collaborative design canvas drops to zero. Are you building software to help people work, or software that does the work?

NTT DATA just surveyed 5,000 enterprise leaders. 60% say cross-border data restrictions are blocking their AI adoption. The bottleneck isn't model intelligence. It's jurisdiction. Are you building your AI strategy for a borderless internet, or a fragmented one?

Cerebras just hit a $95 billion valuation on $510M in revenue. Nvidia makes $510M in about two days. The market isn't pricing Cerebras on sales. They are pricing them on desperation. Enterprise buyers are terrified of an Nvidia monopoly, and Cerebras is selling the insurance policy.

Gartner just surveyed 350 large enterprises deploying AI. 80% of them cut jobs. The financial return on those cuts? Zero. Klarna replaced 700 CS roles with AI, watched quality decline, and started rehiring. Are you deploying AI to replace your team, or to make them better?

OpenAI just launched a $4 billion consulting firm. They are quietly admitting that the model alone is not enough. The hardest part of enterprise AI is not the reasoning engine. It is the execution layer. Are you buying AI models hoping they work out of the box, or investing in the execution layer?

Alphabet just passed Nvidia to become the world's most valuable company at $4.8 trillion. Everyone thought Google was the loser of the AI boom. But the market just realized something fundamental. Having the smartest model doesn't matter if you don't own the infrastructure to run it. Are you building around the smartest model, or the most reliable infrastructure?

Elon Musk called Anthropic "evil" three months ago. Now SpaceX is taking $4 billion to lease them 220,000 GPUs. Principles have a price tag. In the AI arms race, compute is the only currency that matters. Are you building around vendor loyalty, or just buying compute?

The companies building the future of AI can't even support their own customers. Try getting a human on the phone at OpenAI. AI margins die at the prompt when you're running LLMs for customer service without an execution layer. Are you buying AI that just answers questions, or AI that actually resolves tickets?

Google, Microsoft, and xAI just agreed to let the US government test their AI models before release. Silicon Valley spent two years complaining that European AI regulation was stifling innovation. Now the US is quietly building the exact same oversight infrastructure. The US just admitted Europe was right all along.

Anthropic just launched a $1.5B consulting firm with Blackstone and Goldman Sachs. For every $1 on software, companies spend $6 on services. The big consulting firms owned that market for decades. The future of AI revenue isn’t software licensing. It’s consulting, rebuilt from the model up.

Big Tech just projected $674 billion in AI spending for 2026. In the same breath, they laid off 90,000 tech workers. Meta is cutting 8,000 jobs while raising its AI budget to $145 billion. The math is brutal. $674 billion in. 90,000 people out. This isn't just a restructuring. It's a massive reallocation of capital from human talent to compute power.