August 5, 2026
Shopify told TechCrunch this week that AI-driven traffic and orders to its merchant stores tripled year over year in Q2, and crucially, that this traffic is additive rather than a replacement for Google search. That's a genuinely useful data point in an industry that's mostly been running on anecdote. Publishers have spent the last two years watching AI answer engines summarize their content and keep readers from ever clicking through. Shopify's merchants, by contrast, appear to be getting discovered by AI shopping assistants and then still completing a purchase on the actual site. The likely explanation is structural, not magical: e-commerce has a hard conversion event at the end of the funnel that an AI summary can't substitute for. You can read a ten-second AI answer instead of a news article, but you can't wear an AI answer instead of a jacket. That distinction matters for any business trying to figure out where to put its energy. If your value is information, AI search is a threat. If your value is a transaction, AI search might be a new acquisition channel. Businesses running their own marketing and campaigns tooling should be watching referral sources closely right now, because this shift is happening quietly and it's not evenly distributed across industries. I'd still caution against reading too much into one company's self-reported numbers — Shopify has every incentive to frame this positively for merchants — but the underlying logic holds up.
Anthropic is now hiring a team to design its own custom AI chips, co-designing hardware and models together so Claude runs faster and cheaper. This follows the well-worn path OpenAI, Google, and Amazon have already taken, but it confirms something important: the leading AI labs no longer see chip dependency as an acceptable long-term cost of doing business. When you're spending billions on inference and training, shaving even a modest percentage off compute costs through custom silicon is worth the multi-year engineering investment. For business buyers, this is mostly a signal rather than something with immediate impact. Custom chip programs take years to bear fruit, and Anthropic won't be swapping out Nvidia GPUs anytime soon. But it does suggest the economics of frontier AI are tightening, and labs that control more of their own stack will eventually be able to price more aggressively — or pocket more margin. Either way, it's a reminder that the AI vendors your business relies on are themselves locked in a capital-intensive arms race, which is exactly why we've argued that betting your operations on a single model provider's roadmap is riskier than building on a flexible platform. It's also worth noting Nvidia isn't sitting still either — the chipmaker is already pushing security proposals through the week-old Open Secure AI Alliance it helped form, which now counts over 120 companies. That group is worth watching for anyone thinking seriously about agent security as autonomous AI agents become normal parts of business workflows.
MacPaw's move to offer developers on-device inference through Liquid AI's models, so its Eney assistant can run locally instead of calling out to a cloud API, is a smaller story but part of the same pattern. Running AI locally cuts latency, cuts cost, and sidesteps privacy concerns that come with shipping user data to a third-party server. Expect more app makers to quietly follow this playbook as on-device models get good enough for narrower tasks. It won't replace cloud-scale models for complex reasoning, but for a huge share of everyday assistant tasks, it doesn't need to. Taken together, these three stories point to the same underlying trend: the AI industry is maturing past the model layer and into infrastructure, distribution, and unit economics. That's a healthier phase for the technology, even if it's less flashy than a new chatbot demo.
If AI search is already reshaping how customers find your business, are you tracking where that traffic is coming from — or still just watching your Google numbers?
Sources