Pace

Sustainability US Equity Research 13 September 2026

George Gianarikas | Analyst | Canaccord Genuity LLC (US) | | Jason Tilchen, CFA | Analyst | Canaccord Genuity LLC (US) | | Matt Paciulli | Associate | Canaccord Genuity LLC (US) | |

Pace "When trillions of dollars hang in the balance, real commitments around data centers and power, we believe that velocity of spillover is everything. Everything. Let's say it again, that velocity, that pace is important." - From our "Let them eat cake" note, 8/28/26

It is, quite simply, the whole of it. US economic growth now swings on what feels like a single hinge: the pace of the AI buildout. In the first half of 2026, roughly 70% of American growth by our math rested entirely on that single pillar.

Which renders Anthropic CEO Dario Amodei’s weekend declaration perhaps the most consequential pivot in the modern history of the American economy.

Beyond the recent GDP metrics, we have all grown familiar with the sheer scale of this infrastructure wave, a buildout whispered in the same breath as the transcontinental railroad and the dawn of the internet. It is another rare American hour when the nation reaches for its picks and axes, rises to its full height, and hammers out the architecture of a new world.

And—we know the script by heart. First comes the grand design and the surge of hope, then the collective vertigo of a market swollen on its own momentum. We chase the future with a relentless hunger, mistaking motion for destiny. But the hardest truths belong to the morning after when the frenzy subsides, the lights come up harsh and sudden, and we are left to sweep up the broken glass of our own magnificent expectations.

Do Mr. Amodei’s weekend musings (with which both Open AI CEO Sam Altman and SpaceX/Tesla CEO Elon Musk agreed) mark the dawn of the morning after for AI infrastructure? They just might. After all, those three companies anchor the majority of the AI infrastructure buildout, based on our estimates.

So what did Mr. Amodei say? We have all likely read it by now, but here is a brief synopsis of his post: Dario Amodei argues that the AI industry is rushing headlong into a future it can no longer control driven by recursive self-improvement with recent tests showing alarming autonomy.

(Recursive self-improvement is a process where an AI system analyzes, rewrites, and upgrades its own source code/architecture to become smarter. Once it becomes more capable, it can make even better improvements to itself, which can then trigger a continuous, accelerating loop of intelligence growth.)

Mr. Amodei argues that “we must slow the pace at which we improve the capabilities of AI models.” Deceleration buys time for safety research.

Canaccord Genuity is the global capital markets group of Canaccord Genuity Group Inc. (CF : TSX) The recommendations and opinions expressed in this research report accurately reflect the research analyst's personal, independent and objective views about any and all the companies and securities that are the subject of this report discussed herein.

For important information, please see the Important Disclosures beginning on page 5 of this document.

Sustainability Industry Update

Amodei outlines three solutions. First, independent evaluators need internal lab access. Second, democratic nations must coordinate safety standards. Third, global controls must prevent reckless competition. He concludes that slowing progress can help ensure our survival and that “we owe it to humanity to try”.

His essay also uses the word "pace" 12 times. 12.

Amodei’s framework calls for a turn from speed to “…building AI at a balanced rate…”. By exchanging headlong expansion for disciplined governance, its three pillars (embedded evaluators, democratic coordination, and global pacing) could flatten the growth curve of compute and AI-driven energy demand. Labs may lengthen training cycles and stagger capex, replacing endless waves of mega-clusters with more deliberate growth. State-backed coordination and compliance requirements could add real friction to hardware deployment. The proposal swaps an unchecked, hyper- exponential race for a steadier course.

So, a cold dawn could be breaking on the AI…