Guide to the AI barnyard: mapping every major AI player as a farm animal

Summary

The slop trough (Meta). Companies churning through data without clear differentiation are pigs eating slop. Meta fits here, along with anyone producing undifferentiated AI. One counterpoint is that slop today becomes quality product tomorrow, and pig farming can be profitable. In a few years, this slop won't be slop anymore.

The fox circling the henhouse (Oracle). Oracle is the henhouse and the fox is circling but hasn't fully entered yet. The metaphor captures the risk to legacy enterprise software vendors from AI-native competitors. The chickens don't seem panicked yet. Companies with strong cash flow know they need to guard their data and customer relationships.

The cash cow (NVIDIA). NVIDIA was making more than 100% of AI industry profits while everyone else lost money. Even now, with TPU competition from Google starting to show cracks, NVIDIA remains dominant. NVIDIA's recent earnings beat didn't prevent a stock selloff, suggesting market exhaustion rather than operational weakness.

The bull in the china shop (Elon/xAI). Elon disrupts in two ways. He's building data centers in eight months that would take others eighteen, and he's shipping features faster than established players can discuss them. He moved into AI companion territory, an unspoken boundary most frontier labs respected until now.

Lipstick on a pig (Apple Intelligence). Apple's privacy-first ecosystem made it hard to deploy commodity AI features at launch. Unlike Google or Meta, which trained on user data freely, Apple had built a brand around not doing that. The result felt like marketing layered onto a shallow integration.

The dark horse (Thinking Machines). Mira Murati's post-OpenAI startup has blog posts, a product for RL fine-tuning, and is raising capital. It's more visible and well-capitalized than a typical dark horse. It's a bright horse.

Gift horse being looked in the mouth (public market investors). Public investors are selling stocks despite AI beating all expectations. They should be grateful for the AI megatrend and buying more. Instead, they're scrutinizing CEO plans and questioning unit economics.

Workhorse (Amazon). Amazon is reliably dragging the plow without overextending. It's building data centers for Anthropic, hosting models, and staying out of flashy bets or consolidation deals. Not fast-moving, not aggressive, just steady. Amazon hasn't signed the rumored OpenAI-to-commerce partnership yet.

Black sheep/black wolf (Andrej Karpathy). Karpathy broke rank from the AI-acceleration consensus, arguing on Dylan Patel's show for a decade of agents rather than near-term AGI. He popped the bubble at a controversial moment, despite coming from OpenAI and Tesla. The dissent cost him reputation points in a moment when staying with the narrative is rewarded.

The elephant in the room. The $1.4 trillion question is how labs will actually spend that much capital and over what timeline. It's the unspoken tension in every financing round. You never ask a lab founder how they'll deploy or pay for that kind of capital.

Bird's eye view (Leopold/Situation Awareness). Leopold has panoramic awareness of the board and is positioned well.

Lion's share (Satya Nadella/Microsoft). Nadella extracted the lion's share from the OpenAI partnership: 20% of upside, IP rights, control over chips, and dominance of the contract terms. It's an incredible deal-making performance.

Sitting duck (Reddit). Reddit didn't understand how valuable its training data was when it signed the OpenAI deal early. Every model company has now trained on Reddit's corpus. It got caught off guard and lost leverage. Reddit's market cap has climbed to roughly parity with CoreWeave, despite what feels like a raw deal on data licensing. Reddit was sold to Condé Nast for $10 million, spun out, languished in the hundreds of millions, and is now a multibillion-dollar asset.

Headless chicken (Perplexity). Perplexity has multiple, sometimes contradictory strategies. It's building a browser, a Bloomberg terminal, a financial site, trying to acquire TikTok and Chrome, and launching a venture fund. Perplexity's browser gets independent praise, but the company's lack of focus makes it hard to take seriously. Consensus is short Perplexity.

Snake in the grass (Chinese open source models). DeepSeek, Kimi, and other Chinese open-source models are improving rapidly and gaining traction, particularly on price. One view holds they're a stalking horse for negotiating price with Western closed-source models, similar to how Linux functioned. Another worries about soft power and geopolitical bleed-through if these models proliferate globally. The counterargument is that most business use cases like fraud detection and grammar checking are commoditized and won't care about embedded cultural values. Fine-tuning can remove bias. The disagreement persists on severity.

Early bird that got the worm (Josh Kushner). Kushner was very early to OpenAI, doubled down multiple times, and captured enormous upside.

Donkey work (Anthropic). Anthropic is doing the laborious, unglamorous work. Coding, APIs, enterprise sales, entry-level white-collar automation. Not flashy. Not consumer. Not doing a browser or playing for narrative dominance. This positioning is more honest than the noise around competitor launches, though Anthropic also gets framed as the scapegoat in some circles.

Fat cat (Google). Google is rich in every dimension: cash flow, TPUs, researchers, talent. It's releasing models constantly and leads on multiple benchmarks. It has the best image model and top text models depending on the metric.

Golden goose (absent). The golden goose was the neo-cloud and infrastructure play, but the market overheating has killed the gains. CoreWeave still lays golden eggs through GPUs and subsidized data center capacity, but stock appreciation stalled. The metaphor invokes the fable where greed kills the source of sustainable profit.