Avalerion Intelligence · Edition 020
Tuesday 20 October 2026 · AI, Data & Enterprise Transformation for the Nordics, UK & Ireland
Meta’s new terms reach your website on 30 October
This week at a glance
| Move | Who | Why it matters to you |
|---|---|---|
| New business terms let Meta crawl and train on websites linked to any business using its tools | Meta (in force 30 October) | Your marketing team may be accepting a data licence over your website |
| Gemini 4 Argon generally available at $2 and $10 per million tokens until 30 November, then $4 and $20 | Google (13 October) | Business cases built on launch pricing double on 1 December |
| Muse Wonder, Meta’s first frontier model, at $4 and $20 per million tokens | Meta (12 October) | Frontier list prices are converging |
| Q3 revenue $46.72 billion, up 41.2 percent; Q4 guided to $52.3 to 54.3 billion | TSMC (15 October) | AI compute stays a seller’s market into your 2027 budget |
| Two frontier AI pilots helped fix over a thousand potential vulnerabilities in government systems | UK government (12 October) | Using AI on your own defences is becoming the baseline |
| $6.3 million for an engineering firm that sells outcomes, not hours | Solvera, Oslo (15 October) | Agent-run services will reprice what you buy by the hour |
| “Component costs are a headwind to gross margin next year” | Ericsson (15 October) | Expect the pressure in 2027 hardware and network renewals |
1. Top AI News
The frontier price list has a cliff in it.
What happened. On 13 October Google made Gemini 4 Argon generally available through the paid Gemini API, AI Studio and eligible Gemini Enterprise subscriptions, at an introductory $2 per million input tokens and $10 per million output tokens. After 30 November the price rises to $4 and $20. A day earlier Meta released Muse Wonder, its first frontier model, at $4 and $20 through its API.
Why it matters. Two frontier models now list at the same price, and one is half price for six more weeks. Pilots costed this month will look twice as cheap as production.
What to do. Cost every 2027 business case at list price, not launch price, and put the end date of any promotion into the plan.
The chips behind it stay scarce.
What happened. On 15 October TSMC reported third-quarter revenue of $46.72 billion, up 41.2 percent in dollars, and guided the fourth quarter to $52.3 to 54.3 billion. On the call, chief executive C.C. Wei said AI demand is still well above what TSMC can supply.
Why it matters. When the world’s chipmaker cannot meet demand, compute stays a seller’s market, and cloud and model prices follow it.
What to do. Treat AI capacity as a constrained input in 2027 planning: commit what is proven, and keep experiments on flexible terms.
2. Enterprise AI Trend: AI turned on your own defences
The UK uses frontier models to find its own holes.
What happened. On 12 October UK AI minister Kanishka Narayan said in a speech at Imperial College London that two government pilots using frontier AI have helped identify and fix over a thousand potential cyber vulnerabilities across government systems. The government is preparing for serious incidents involving autonomous AI systems, and his message to developers was blunt: “if you cannot adequately control the risks, do not release the system.”
Why it matters. Attackers have the same models. Finding your own weaknesses with AI is becoming a basic expectation, not an innovation project.
What to do. Ask your security lead whether AI-assisted vulnerability scanning runs on your own estate, and who owns the incident plan if one of your agents misbehaves.
3. Data & AI Readiness: your website is an AI input now
Meta writes AI training into its business terms.
What happened. Meta’s updated Commercial Terms take effect on 30 October. A new paragraph authorises Meta to use automated software, including crawlers, to collect content from the domains and web pages associated with a business’s products or services, and to use it to develop, research and train its products. Legal experts consulted by Nieman Lab read it as broad enough to cover news sites. Meta calls those reports false and says there is no change to how it uses web crawlers. There is no partial opt-out: rejecting the terms means no longer using Meta’s business products.
Why it matters. Whoever runs your Facebook Page or ad account may be accepting a data licence over your website. Legal rarely sees that click.
What to do. Before 30 October, have legal read the new terms, check robots.txt on every domain you care about, and decide which content you are willing to license this way.
4. Business Process Spotlight: selling outcomes, not hours
An Oslo engineering firm runs on agents.
What happened. On 15 October Oslo’s Solvera announced a $6.3 million pre-seed round led by Seedcamp. It is an engineering firm of record for construction, starting with mechanical, electrical and plumbing design. AI agents handle a growing share of the work and human engineers make the judgment calls. It sells services, not software licences or billable hours, and already works on Norwegian building projects, including for contractor Betonmast.
Why it matters. When agents do part of the work, the hour stops being the unit of value. Your suppliers will start pricing on outcomes, and your own service teams may have to.
What to do. Pick one service you buy by the hour and ask the supplier what it would cost as a fixed deliverable.
5. Nordic Technology: Ericsson feels the component bill
AI efficiency arrives as input costs rise.
What happened. On 15 October Ericsson reported third-quarter sales of SEK 59.5 billion and 7 percent organic growth. Chief executive Per Narvinger said customers are already seeing efficiency gains from AI-enabled features in their networks, warned that “component costs are a headwind to gross margin next year”, and announced a simpler operating structure.
Why it matters. A Swedish hardware maker is warning of dearer components in the same week TSMC said it cannot meet AI demand. Expect that pressure in your own 2027 renewals.
What to do. Ask your hardware and network suppliers now how component prices will move in 2027 contracts, before the renewals land.
Join us in Stockholm
Thursday 5 November 2026, 12:00 to 13:30, Klarabergsviadukten 63, 111 64 Stockholm. Free.
A working lunch with your laptop open: you build a real lifecycle flow in Customer.io, guided by our team and Customer.io’s, and leave with it working in your own environment. Not a demo, not a panel. Seats are kept small.
Do this next. Thirty minutes with us: which AI costs in your 2027 plan rest on launch pricing, which terms your teams have accepted on the company’s behalf, and what your website and data are licensed for. No pitch, no deck.
Recommended reading: What AI readiness actually means for Nordic companies and Why most enterprise AI projects still fail, at avalerions.com/insights
Frequently asked questions
What changes in Meta’s Commercial Terms on 30 October 2026?
The updated terms add a paragraph authorising Meta to use crawlers and other automated software on the domains and web pages associated with a business’s products or services, and to use that content to develop, research and train its products. They apply to businesses using Meta’s commercial products. Meta says the update does not change how it uses web crawlers.
How much does Gemini 4 Argon cost?
Google made Gemini 4 Argon generally available on 13 October 2026 at an introductory $2 per million input tokens and $10 per million output tokens, with cached input 95 percent off. After the introductory period ends on 30 November 2026, prices rise to $4 per million input tokens and $20 per million output tokens.
What did TSMC report for the third quarter of 2026?
On 15 October 2026 TSMC reported third-quarter revenue of $46.72 billion, up 41.2 percent year on year in US dollars, with a gross margin of 66.4 percent. It guided fourth-quarter revenue to between $52.3 billion and $54.3 billion, and said AI demand remains well above what it can currently supply.
What is Solvera?
Solvera is an Oslo-based, AI-native engineering firm for construction. It acts as engineering firm of record, starting with mechanical, electrical and plumbing design, using AI agents for a growing share of the work and human engineers for judgment. On 15 October 2026 it announced a $6.3 million pre-seed round led by Seedcamp.
Sources
- Meta, Commercial Terms (preview of the version effective 30 October 2026): facebook.com
- Nieman Lab, “Meta’s new commercial terms will allow the scraping of news sites, legal experts say”, October 2026: niemanlab.org
- PPC Land, “Meta’s new terms let it crawl business websites to train AI from Oct 30”, 16 October 2026: ppc.land
- Google, “Gemini 4 Argon: Built to solve your hardest problems”, 13 October 2026: blog.google
- TechCrunch, “Meta releases its first frontier model, Muse Wonder”, 12 October 2026: techcrunch.com
- TSMC, third quarter 2026 earnings release (Form 6-K), 15 October 2026: sec.gov
- CRN Asia, “TSMC third quarter revenue tops guidance as CEO says AI demand still exceeds supply”, 15 October 2026: crnasia.com
- Imperial College London, “AI minister announces measures to protect against AI risks in speech at Imperial”, 12 October 2026: imperial.ac.uk
- Tech.eu, “AI minister calls on AI frontier labs to ramp up safety, as UK ups cyber defences”, 12 October 2026: tech.eu
- Dealroom, “Oslo’s Solvera raises $6.3M to become an AI-native engineering firm”, 15 October 2026: dealroom.co
- Ericsson, “Ericsson reports third quarter results 2026”, 15 October 2026: ericsson.com
- Event page (re-checked 19 October 2026): avalerions.com
Avalerion Intelligence, AI, Data & Enterprise Transformation for the Nordics, UK and Ireland. Published by Avalerion Consulting AB, Stockholm. Subscribe by email