The Week AI Got Cheaper, Bigger, and More Practical
Last week was one of those weeks where it felt like every major AI lab had something to announce. Some of it was hype. Some of it was genuinely interesting. And buried in the headlines were a few shifts that small and mid-sized businesses should actually pay attention to.
Here's what happened — and what it means for the rest of us.
1. The Models Got Bigger (Again)
Anthropic released Claude Mythos 5, the first widely recognised ten-trillion-parameter model, pitched at high-stakes work like cybersecurity, research and complex coding. OpenAI shipped GPT-5.5 — positioned as a step toward a unified "super app" that bundles chat, coding tools and browser capabilities into a single interface. Alibaba pushed out Qwen 3.6-Max-Preview, with strong jumps in coding and agentic performance.
For most businesses, the headline numbers don't matter. What matters is that the gap between what a frontier AI model can do and what's useful for a normal business is narrowing every quarter. Tasks that almost worked six months ago are now boringly reliable.
If you tried automating something with AI a year ago and it didn't quite work — it's worth trying again.
2. The Models Also Got Cheaper
DeepSeek shipped an open-source model that scores within striking distance of the leading commercial models — at roughly one-sixth the inference cost.
This is the more important story for SMEs.
The cost of running AI is collapsing. The same task that cost a dollar to automate eighteen months ago might cost a few cents today. That changes which problems are worth automating. Things that were uneconomical at scale — every email triaged, every document summarised, every transaction reviewed — are now well within reach.
And open-source frontier models mean you no longer have to bet your business on a single vendor. The optionality is real.
3. The Money Is Real (And Reshaping the Industry)
Google announced an investment of up to $40 billion in Anthropic — $10B in cash now at roughly a $350B valuation, plus up to $30B more tied to milestones, alongside 5 gigawatts of compute over five years starting in 2027. Anthropic and Amazon also expanded their existing partnership for another 5 GW. Anthropic and NEC announced a collaboration to build Japan's largest AI engineering workforce. Meta is roughly doubling its AI capex to $115–135 billion in 2026.
These are not "experimenting with AI" numbers. These are "rebuilding the global compute layer" numbers. The infrastructure being built now will define what's possible for the next decade.
For business owners, the practical implication is simple: AI capabilities are going to keep improving rapidly. Any decision you make today about AI shouldn't assume the technology stays where it is.
4. The Workforce Shifts Are Starting
Snap CEO Evan Spiegel announced the layoff of around 1,000 employees, citing rapid advancements in AI and projecting over $500M in annualised savings. Meta confirmed 8,000 layoffs — about 10% of staff — plus 6,000 unfilled roles, with cuts beginning May 20. Both companies are reallocating those budgets toward AI infrastructure.
This isn't a small-business story directly — but it's a signal. Larger companies are starting to make hard decisions about what work AI can absorb. SMEs don't usually have the same dynamic, but the same underlying question applies:
“<em>Which tasks in your business are repetitive enough that they shouldn't require a human to do them?</em>”
The answer is usually more than people expect.
5. Regulation Is Coming Into Focus
The EU AI Act's main enforcement window opens on August 2, 2026 — about 99 days away. That's when Article 50 transparency rules, the Annex III high-risk regime, and the full penalty structure (fines up to 7% of global turnover) all take effect.
If you operate in or sell into the EU, this is a deadline worth understanding. For most SMEs the obligations are lighter than the headlines suggest, but a quick review of how you're using AI — and how you're disclosing it — is sensible homework for the next few months.
What This Means for Your Business
A few practical takeaways from the week:
- The "wait and see" window is closing. AI is no longer a frontier experiment. It's a standard line item in operating budgets at every level.
- Cost is no longer a real barrier. The most common reason SMEs delay AI adoption — "it's too expensive" — is becoming less true every month.
- Vendor lock-in is getting easier to avoid. Open-source models give you options that didn't exist a year ago.
- Start small, start now. The businesses getting value from AI today aren't the ones doing the biggest projects. They're the ones who picked one repetitive task, automated it, and built from there.
At The Empyrean, we help businesses cut through the noise around AI and connect it to the systems they already have — without the overhaul. If you'd like an honest read on where AI could actually help your business, we're happy to take a look.