The 10% That Changes Everything
Why “close enough” built the consumer wave but won’t build the enterprise one
Anyone who uses AI regularly has experienced this. It gets most of the way there, then slips on an important detail. It is useful, but not something you would trust without checking. Getting 90% right can feel like magic, but not dependable enough for work with no room for error.
This morning Jensen Huang explained why that last 10% is about to matter a lot more. He was on CNBC at the Synopsys AI Engineering Conference talking about NVIDIA’s new $2B stake in Synopsys, and the conversation quickly shifted to a bigger point. Consumer AI was built on “close enough.” Enterprise AI cannot be.
Here is what Jensen said
“The important concept here is accelerated computing and AI is revolutionizing every single industry. Of course it started in the consumer segment, because getting every answer 90% right is a pretty magical thing.
But in the industrial segment, that 10% you don’t get right becomes mission critical. And so it takes longer for us to create the tools, create the technology necessary, for the industrial sectors, the enterprise segments, to be able to adopt this technology. But the time is now here.
[…] During 2016 the world’s scientific supercomputers was 90% CPUs and 10% GPUs. This year it’s 90% GPUs and 10% CPUs. The platform shift has happened.
What Synopsys does in engineering is founded on principled physics. We are now making this shift in this industry. It is one of the most computing intensive industries in the world, and we are elevating it with AI.”
This matters for a few groups
1. Consumers. Your AI tools will get more reliable as enterprise standards force accuracy higher. Better grounding. Better context. Less drift.
2. Operators and founders. The last 10% is where real enterprise value is. This is the shift from fun demos to workflows that actually run the business.
3. Enterprise buyers. Once the hardest scientific workloads flip to GPUs, the rest of the economy follows. You are stepping into a new compute baseline whether you planned for it or not.
4. Investors. The consumer phase created demand. The enterprise phase creates margin. Precision AI is where durable economics show up.
5. Engineers and product teams. Everything from chip design to supply chain planning is about to compress cycles and rebuild itself around accelerated computing.
How I’m looking at this
The remaining 10% is where the risk and opportunity sits. It is where regulation lives. It is where trust either forms or breaks. It is also the part “There Is No Magic Bullet” calls out, the part that never gets solved by one breakthrough.
It also ties back to what I wrote in “Billions of Small Minds” (here). The real advantage goes to leaders who know where their business sits on the adoption curve. Not the hype curve. The adoption curve. Some companies can live with “close enough” for a while. Others can’t. Some markets move fast. Others punish mistakes. Timing matters more than enthusiasm.
Seen this way, the 10% stops looking small. It becomes the real frontier. Planes, hospitals, factories, and banks operate there. Almost right is still wrong in those environments.
If 90% is good enough for your business, move fast and experiment (my Speed to Value Matrix can help).
For most companies it is not. Not without guardrails. Not without training. Not without people who know how to use these tools with discipline. Knowing where you sit on that curve is becoming a leadership skill.
Scott