We’re watching something bizarre unfold in Silicon Valley right now. Every major tech company is throwing ungodly amounts of money at AI models, each one claiming theirs will be the breakthrough that changes everything. Spoiler alert: most of them won’t.
The pattern is getting predictable. A company announces a new large language model with some incremental improvement, tech journalists scramble to test it with creative writing prompts, and within weeks everyone’s moved on to the next announcement. It’s exhausting and expensive, and I’m not sure anyone outside the boardrooms really understands why we’re doing this anymore.
The Billion Dollar Question
Here’s what really gets me. These models cost hundreds of millions to train. Google, Microsoft, Meta, they’re all burning cash like it’s going out of style. And for what? So their chatbot can write slightly better poetry than the last version? Don’t get me wrong, the Technology itself is impressive. The engineering is remarkable.
But impressive engineering doesn’t always equal useful product. Remember when every company needed to have a blockchain strategy? This feels eerily similar, except the stakes are way higher because the infrastructure costs are astronomical.
The energy consumption alone should make us pause. Training these massive models requires data centers running at full capacity for months. Some estimates suggest training a single frontier model uses as much electricity as a small city consumes in a year. That’s not sustainable, and it’s definitely not something we should be doing just because everyone else is doing it.
When Competition Becomes Redundancy
The real problem is that most of these models are solving the same problems in nearly identical ways. Sure, one might be 3% better at coding tasks while another excels at translating obscure languages. But do we really need fifteen different companies each spending a billion dollars to achieve basically the same thing?
This is where the Business logic falls apart for me. In a rational market, companies differentiate or die. But in AI right now, everyone’s convinced that being second place means becoming irrelevant. So they’re all building the same thing, hoping their version somehow becomes the standard.
OpenAI releases GPT-5, Google counters with Gemini Ultra 2.0, Anthropic drops Claude 4, Meta open sources Llama 4. The cycle repeats every few months. Meanwhile, actual problems that AI could help solve get less attention because they’re not sexy enough for the headlines.
The Innovation Theater Problem
What bothers me most is how much of this feels performative. Companies aren’t necessarily building these models because they have a clear use case or customer need. They’re building them because not building them would signal weakness to investors and competitors.
That’s not innovation. That’s just expensive theater with really good special effects.
Some of the most interesting AI applications I’ve seen lately come from smaller companies using existing models in clever ways. They’re not trying to reinvent the wheel, they’re just finding better ways to use the wheels we already have. But those stories don’t generate the same buzz as “Company X announces $10 billion investment in next-gen AI.”
The venture capital money keeps flowing though. Investors are terrified of missing the next big thing, so they’re funding everything with “AI” in the pitch deck. This creates a feedback loop where companies keep building bigger models because that’s what gets funded, not because it’s what the market actually needs.
Where Do We Go From Here
I’m not arguing that AI research should stop. Far from it. But maybe we need fewer frontier models and more focus on making existing ones actually useful, efficient, and accessible. Maybe we need to ask harder questions about energy use and environmental impact before training another parameter-bloated beast.
The funny thing is, in five years we’ll probably look back at this period and wonder what we were thinking. Either one or two models will have won and the rest will be forgotten, or we’ll realize we were solving the wrong problem entirely and pivot to something else. Either way, a lot of smart people will have spent a lot of time and money chasing something that in retrospect seemed obvious to avoid.
But that’s the nature of innovation, I guess. Sometimes you have to watch everyone make the same expensive mistake before you figure out the right path forward. The question is whether we can afford to keep making this particular mistake at this particular scale, or if we’ll wise up before the bill comes due.
