This Week in Tech: Cheaper AI Chips, a Talent War, and What It Means for Your Service Business

If you run a service business, you don't spend Monday morning reading semiconductor press releases — you're staring at a full inbox, a half-staffed schedule, and quotes you haven't sent. So here's the deal we make every week at PointWake: we read the tech news so you don't have to, and translate it into the one thing that matters — whether it changes how you should run your shop. This week the companies building AI quietly moved to make it dramatically cheaper, the people who build the smartest models started playing musical chairs, and the tools that let small teams automate real work kept shipping. Here's what happened and why it lands on your desk, not just in a data center.

On June 24, OpenAI and Broadcom revealed Jalapeño, OpenAI's first custom-designed AI chip, built specifically for 'inference' — the everyday work of answering prompts, not training a model from scratch. Broadcom's CEO said the chip shows cost savings of roughly 50% compared with the typical AI GPUs most providers rely on today, and it went from design to manufacturing in about nine months. Why should a plumbing company, HVAC outfit, or law office care about a chip? Because the cost of running AI is the cost you eventually pay. Every AI scheduling assistant, automated quote-writer, and after-hours chatbot runs on inference. When the underlying compute gets cheaper, the tools built on top of it get cheaper, faster, or both. AI features that felt like an expensive luxury six months ago are sliding into no-brainer territory. If you held off automating because the math didn't work, the math is changing under your feet.

The same week, TechCrunch reported that AI researchers keep leaving Google for rivals. Noam Shazeer — a co-author of the 2017 paper that introduced the architecture behind nearly every modern AI model — is heading to OpenAI, and two more senior Gemini researchers are joining Anthropic, making four senior Google exits in roughly six days. The signal for a small business owner is real: the AI platforms are in a knife fight for talent, which means they're in a knife fight for you as a customer. That competition is exactly why prices keep dropping and free tiers keep getting more generous. It also means no single vendor is a safe forever-bet. This is why PointWake's audit-first approach never hard-wires you to one AI provider — we build your workflows so the assistant doing your intake or follow-ups can be swapped as the market shifts. You own the process, not a logo.

Reporting this month confirmed that Apple's upcoming iOS 27 will introduce an Extensions system that lets users choose which AI powers Siri — Claude, Google Gemini, or ChatGPT — instead of being locked into one. The lesson for a service business isn't about Siri; it's the principle: the winning setup lets you choose the right tool for each job instead of forcing everything through one app. Most owners we audit are doing the opposite — a CRM that half-does scheduling, an email tool that half-does invoicing, and a phone that doesn't talk to either. The future Apple is signaling is modular: pick the best assistant for intake, the best for follow-up, the best for quoting, and wire them together. That's the architecture PointWake builds toward.

On June 2, GitHub made its Copilot SDK generally available, shipping the same agent engine behind Copilot across six programming languages with support for connecting outside tools. In plain English: developers can now drop an AI agent — software that plans and executes multi-step tasks on its own — into their own products. The broader June launch wave included business tools for automating routine cross-app workflows like reporting, scheduling, and customer communications. For two years, AI mostly meant a chatbot that gave you a suggestion you still had to act on. The agentic wave means software that completes the loop — it reads the new lead, drafts the reply, books the slot, and updates the record without you babysitting it. LinkedIn's own hiring data this spring pegged time savings from a single AI tool at about 5.1 hours per week. Stack a few workflows and you're reclaiming a workday.

For context on how fast this is moving: Anthropic, the maker of Claude, confidentially filed to go public this month, with reporting putting its revenue run-rate at roughly $47 billion in May 2026, up from about $10 billion a year earlier. You don't need to care about the IPO. You need to understand what that growth funds: relentless improvement in the exact tools your business can use. The roadmap isn't slowing down — it's accelerating, and it's pointed squarely at making small teams operate like big ones.

Put the week together and one trend emerges: AI is getting cheaper to run, more competitive to buy, more modular to assemble, and more capable of doing real work end-to-end. For a service business, that's not a reason to panic about robots — it's an invitation to fix the boring, expensive bottlenecks (missed after-hours calls, quotes that sit for three days, follow-ups that never go out) for a fraction of what it cost last year. But cheaper tools only help if pointed at the right problem. The most common mistake we see is owners bolting AI onto a broken process and automating the chaos. That's why everything at PointWake starts with an audit: we map where your time and leads actually leak before we automate anything, then build workflows — intake, CRM, scheduling, follow-up — that run whether or not you're holding your phone. Your action item this week: pick the one task you most dread doing manually, write down how many hours it eats per week, and make that number your starting line. The tools to erase it just got cheaper — and they'll be cheaper again next quarter. Want to know exactly where automation would pay off fastest in your business? That's literally what we do. Free AI Readiness Consult → https://pointwake.com/contact