On August 1, OpenAI announced that an internal version of its Astra model solved ten open problems across mathematics and theoretical computer science. Not exam questions. Not textbook exercises. Open problems that had resisted human mathematicians for decades, including a construction proving the existence of non-sofic groups, one of the central unresolved questions in group theory. The model published formal proofs on GitHub. Fields Medal winner Timothy Gowers said he would recommend one of them for publication in a top journal without hesitation.
The compute cost was roughly two thousand dollars.
Now hold that number in your head while you read the next one. BCG published its 2026 analysis of AI-first cost advantage earlier this year. Sixty percent of companies report minimal or no value from their AI investments. Not “disappointing returns.” Minimal or none. These are organizations that committed budget, hired consultants, ran pilots, bought licenses, stood up internal teams, and got functionally nothing back.
Two thousand dollars solved what Fields Medalists could not. Millions bought silence.
The gap is not about capability
This is the part most coverage misses. The Astra story is being reported as a science breakthrough. It is. But for anyone running a business, the real story is what it reveals about distance.
The frontier of what AI can do moved again last week. It moved into territory that required decades of specialized human training to even attempt. And it did it for less than a month of one employee’s salary.
Meanwhile, BCG found that in a typical AI implementation, only 10% of the value comes from the algorithm and 20% from the technology and data. The remaining 70% comes from managing process change. Seven out of ten dollars of potential value sit in how an organization decides to reorganize its work. Not in the model. Not in the vendor. In the willingness to change how things run.
The organizations stuck at zero ROI did not pick the wrong model. They refused to change the process around it.
What the leaders are doing with that same capability
BCG’s numbers for companies that do reorganize are not marginal. AI leaders deliver three times greater cost reduction than peers. They produce 1.6 times higher EBIT margins. Their return on invested capital runs 2.7 times higher.
The firms BCG calls “future-built” companies see five times the revenue increases and three times the cost reductions from AI compared to everyone else. These are not pilot results. These are structural advantages built into how the company operates, and they compound every quarter because the organization learns faster with each cycle.
Nobody talks about this part in the Astra coverage. The capability is accelerating. The organizations that have already reorganized around AI are absorbing that acceleration. They do not need to run a six-month evaluation cycle when something like Astra ships. They have teams structured to integrate new capability within weeks, because the roles, the workflows, and the decision rights were rebuilt to accommodate continuous improvement.
The organizations that have not reorganized are watching the same announcements from the same distance they were at six months ago. Twelve months ago. The capability got faster. They did not.
The structural nature of falling behind
McKinsey identifies roughly 6% of enterprises as AI high performers, the ones attributing more than 5% of EBIT directly to AI. That is a small group. But the performance gap between them and everyone else is not closing. It is widening, because the advantage is not in having AI. It is in having rebuilt the organization to learn from it continuously.
This is where it compounds. A company that reorganized eighteen months ago has eighteen months of institutional learning baked into its operations. Its teams have discovered workarounds, internalized edge cases, developed judgment about when AI works and when it does not. That cannot be replicated with a Q4 training sprint or a sudden vendor contract.
And Meta narrowed its 2026 capital expenditure range upward to $130 billion to $145 billion, driven by AI infrastructure buildout. The investment flowing into capability is accelerating. Whether AI will be powerful enough to matter is no longer a real question. Astra closed that book. What remains open is whether your organization will be structured to use what is already available, let alone what ships next.
What six more months of waiting costs
If your company deployed AI tools in the last year but cannot point to measurable business outcomes, the diagnosis is specific. The technology works. The math is settled — literally. What is missing is not capability. It is the willingness to change how work gets done.
Every month that passes without that reorganization, the companies that already moved add another month of compounded learning. Their costs keep dropping. Their teams keep getting sharper. And the distance between them and the companies still running evaluation committees grows in a way that money alone cannot close.
Two thousand dollars solved problems that stumped human experts for generations. The cost of organizational inaction is running significantly higher than that, and the invoice comes due a little more each week.