Making Sense.

Nvidia, the company that makes most of the specialized chips powering today's AI systems, announced a deal with a group of major financial firms, including Goldman Sachs and BlackRock, to raise more than $500 billion from investors to fund AI computing infrastructure. This deal aims to turn AI chips into a stable, tradeable financial asset, with BlackRock's CEO Larry Fink comparing the structure to the 1970s, when banks first bundled mortgages into a new kind of security investors could buy into.

Other analysts point to a different historical comparison, the 1870s railroad-bond boom, when a similar wave of borrowing to build infrastructure ended in a financial panic once the debt outpaced actual demand. That comparison reflects concern that Big Tech has already issued $194 billion in bonds this year to fund AI infrastructure, and rating agencies have flagged strained cash flow among the biggest spenders.

Regina-based Max Power Mining discovered a large natural hydrogen deposit at its Lawson site in southern Saskatchewan late last year, drilling Canada's first natural hydrogen well. To speed up further exploration, the company built Maxx Lemi, a proprietary AI tool that processes large geological and geophysical datasets to flag promising drilling targets. Natural hydrogen occurs in underground rock formations and emits only water vapour when burned, unlike manufactured hydrogen, which needs expensive, emissions-heavy industrial production. Max Power says the goal is cutting traditional site-identification timelines by months or years, and it's already used the platform to identify several "lookalike" sites across the more than two million acres of permitted land it holds.

For BC's mining, forestry, and energy sectors, this is a useful example of AI aimed at one specific, expensive bottleneck rather than a broad rollout. Worth watching for any resource-sector organization sitting on large, underused geological or operational datasets of its own.

On August 2, two new obligations under the EU AI Act took effect. AI-generated or AI-manipulated content published into the EU market now needs a machine-readable disclosure mark, chatbots and AI agents must identify themselves as automated, and deepfakes must be flagged (the European Commission published a standard icon set for the labels). At the same time, the EU's AI Office gained formal power to investigate and penalize the model-making labs directly, including OpenAI, Anthropic, and Google. Fines for either obligation run up to €15 million or 3% of global turnover.

The rules split responsibility two ways. If your business runs a customer-facing chatbot, whether you built the model or you're calling someone else's through an API, you're the "provider" of that AI system under the Act, and you're responsible for making sure customers know they're talking to AI. Separately, if you publish AI-generated text specifically to inform the public on a matter of public interest, you need to disclose that, though ordinary marketing copy that gets human review before publishing is explicitly exempt. Model providers like Anthropic carry their own separate obligation to mark generative outputs as AI-made, which is why Claude started embedding invisible watermarks in its text worldwide on August 2.