AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: AI Moves Fast In This 2027 Scenario—Until Compute Becomes Scarce on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get smart everyday buys delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Thorsten Meyer’s November 18, 2026, article presents a possible 2027 scenario in which AI helps build better AI until chip, power and data-centre limits slow progress. The scenario is not a forecast established by evidence: METR’s software-task measure has limits, different fits produce different timelines, and broad workplace effects remain unproven.

Thorsten Meyer published a 2027 AI scenario on November 18, 2026, arguing that AI systems could speed up the development of future systems before hitting limits in chips, data-centre construction and electricity. The article is explicitly a scenario, not a confirmed forecast, and its evidence points to uncertainty about both the pace of progress and how quickly it could affect businesses and workers.

Meyer’s proposed chain has four links: AI systems take on more work involved in developing other AI systems; that work accelerates progress; available computing capacity becomes a bottleneck; and, later, automation in physical settings could help expand capacity and deployment. Meyer calls the later phase “embodied self-improvement”. Each link depends on assumptions, and the article says any one could fail.

The article uses METR’s 80% time horizon as one measure of AI agents’ ability to complete software-related tasks. In the reported results, the measure rose from 0.89 minutes for GPT-4 in March 2023 to 185.9 minutes for an early Claude Mythos Preview in April 2026. METR defines the measure by the human-estimated duration of tasks an agent is predicted to complete successfully four times in five.

Those results do not settle the timing of the scenario. Meyer’s fits to METR measurements produce different estimates: an exponential fit using data from 2024 reaches a 16-hour task horizon in January 2027, while a fit using data from 2019 reaches it in June 2028. A hyperbolic fit reaches infinity in November 2027, but the article says this was the worst-performing of five fits in its backtests. These are the author’s extrapolations, not METR forecasts.

At a glance
analysisWhen: Published November 18, 2026; the propos…
The developmentThorsten Meyer published a scenario-based analysis of possible rapid AI progress in 2027 followed by a physical-compute bottleneck.

Compute Could Set the Pace

The scenario matters because it links AI capability gains to the physical resources needed to train and run systems. If AI development speeds up while chips, data centres and electricity expand more slowly, improvements in software may not translate into equally rapid growth in available AI services. Conversely, a later increase in computing capacity could change how widely those systems are used.

The article cites US data-centre construction spending at an annual rate of $96.5 billion in September 2026, compared with $50.7 billion a year earlier. It also cites Epoch AI’s expectation that power will be the first constraint on scaling and that new power supply takes three to five years from a decision. These figures describe investment and a cited assessment of infrastructure timing; they do not establish when a shortage will occur or how severe it would be.

For workers and businesses, the distinction between capability and adoption is central. The source reports that 18% of US firms used AI in a business function in late 2025 and early 2026, while 2% reported AI-related job cuts. Those measures suggest that stronger benchmark performance alone does not show that AI has already transformed most workplaces.

Products Worth Considering

Dell XPS 13 13.4-Inch Touchscreen Laptop, Intel Core 5 320, 16GB RAM, 512GB SSD

Dell XPS 13 13.4-Inch Touchscreen Laptop, Intel Core 5 320, 16GB RAM, 512GB SSD

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

As an affiliate, we earn on qualifying purchases.

What METR’s Measure Captures

METR is a nonprofit that evaluates frontier AI agents. Its time horizon estimates how long a task, measured by human expert completion time, an agent can complete at a specified reliability. The 80% figure means success is predicted on four out of five attempts; it is not a measure of general intelligence or of performance across all jobs.

METR’s task set primarily covers software engineering, machine learning and cybersecurity. The source notes that METR describes these tasks as cleaner than economically valuable work in real settings. It also says the benchmark has measurement limits: results above 16 hours are unreliable with the current task suite, and human completion times were measured for only five of 31 tasks estimated to take at least eight hours.

Another point of uncertainty is the data series itself. According to Meyer, METR marked its public page as no longer actively updated on September 8, 2026. That makes continued measurement and comparable public data important for judging whether the apparent trend persists. A rising historical curve cannot, by itself, establish the rate of future progress.

“We reach ASI either in 2027 or shortly after. The effect: in two months we see the progress of a whole year, and then further progress until we hit a ceiling in available compute.”

— Thorsten Meyer

The Timeline Is Not Established

The article’s central acceleration claim is not demonstrated by the benchmark figures alone. Meyer says a year of progress arriving in two months would amount to about six times the earlier pace, while his account of METR’s estimate puts Anthropic’s improvement on well-specified tasks at around twice the earlier rate, with a possible range of 1.3 to 3 times. The measures and assumptions are not equivalent, so the figures do not prove that a sixfold acceleration is underway.

The article also reports that OpenAI’s GPT-6.2 Astra was the first externally released model from the company to meet its “High” threshold for AI self-improvement. That is a claim about a company-defined threshold; the source does not provide enough detail here to independently assess what the threshold measures or how it compares with other evaluations. The article says Anthropic, OpenAI and METR did not review or sponsor it, and discloses that AI tools, including Anthropic’s Claude, assisted its research and drafting.

Other unknowns include whether the software-task trend extends to broader work, whether power or another resource becomes the binding constraint, and how quickly new infrastructure can be built and connected. Artificial superintelligence has no agreed definition, and the article does not claim that its projected date or later economic effects have been confirmed.

Signals to Watch in 2027

Meyer says the scenario should be tested against future evidence, rather than treated as settled. Relevant signals include whether updated, comparable evaluations show sustained gains in task duration and reliability; whether those gains extend beyond relatively well-specified software tasks; and whether independent evidence supports faster AI-assisted model development.

Infrastructure data will also matter: data-centre construction, chip availability, power connections and the time required to bring new electricity supply online could show whether compute is becoming a practical limit. On the deployment side, business adoption and reported employment effects would help test whether capability improvements are reaching the wider economy.

The source does not specify a formal next release date for METR’s measurements or a scheduled milestone that would resolve the scenario. For now, the 2027 timeline remains a hypothesis. Its credibility will depend on new measurements and infrastructure developments, as well as evidence that the proposed feedback loop is operating at the pace Meyer describes.

Key Questions

Is the article predicting that AI will reach superintelligence in 2027?

No. Meyer presents 2027 as a scenario based on his view, not as an established forecast. The article also says artificial superintelligence has no agreed definition.

What does METR’s 80% time horizon measure?

It estimates the duration of a task, measured by human expert completion time, that an AI agent is predicted to finish successfully four times out of five. The cited tasks are mainly in software engineering, machine learning and cybersecurity.

Do the METR results show that AI will accelerate sixfold?

No. The sixfold figure is part of Meyer’s scenario for compressing a year of progress into two months. The article reports a separate METR estimate of around twice the earlier improvement rate for Anthropic on well-specified tasks, with a cited range of 1.3 to 3 times.

Why might compute become a bottleneck?

AI systems need chips, data-centre facilities and electricity. Meyer cites an Epoch AI assessment that power could be the first constraint on scaling and that new power supply takes three to five years from the decision. The timing and scale of any constraint remain uncertain.

Has rapid AI progress already caused widespread job cuts?

The source reports that 18% of US firms used AI in a business function in late 2025 and early 2026, while 2% reported AI-related job cuts. Those figures do not establish future effects or show that AI progress has already caused widespread job losses.

Source: ThorstenMeyerAI.com

This content is for general information only and is not financial, tax or legal advice. Consult a qualified professional for decisions about your money.
You May Also Like

7 Best Wireless Smartwatches for Prime Day Deals in 2026

Discover the best wireless smartwatches on Prime Day 2026, including Apple, Garmin, and budget options, with details on features, prices, and suitability.

Why The Computer Science Enrollment Dip Has Renewed AI Job Questions

Preliminary fall 2026 data show computer science enrollment down across college levels, as weak entry-level hiring renews questions about AI’s role.

ChatGPT Rank Monitor: The Key To Better Generative Engine Visibility

A new ChatGPT rank monitor tool aims to help brands track their presence in AI-generated answers, addressing a key gap in current SEO and marketing analytics.

AI Teams’ Guide To This Week’s Security Fixes And Threats

A weekly security roundup covers active Fortinet attacks, Android banking malware, a PDF backdoor and urgent patching decisions.