Will AI "Take Over" by 2030? What the Actual Forecasts Say — and What They Mean for Your Money

You have heard the claim. It gets repeated on the news, in podcasts, at dinner: AI is getting very powerful and could take over by 2030.

It is worth separating what is actually being forecast from what gets repeated about it, because the underlying research is more interesting — and more useful — than the headline.

Here is what the credible published work says, what it does not say, and what a person with a TFSA or a 401(k) should reasonably do about it.

What the forecasts actually say

Anthropic's three scenarios

In September 2026, Anthropic published an economic model projecting US outcomes in 2030 under three different assumptions about AI capability and adoption. Note the obvious: this is an AI company modelling the impact of AI, which is a reason to read it carefully rather than a reason to dismiss it. The model's assumptions are published, which is more than most forecasts offer.

Modest scenario. AI turns out to be roughly as economically significant as the internet — real gains, within historical norms for a major new technology. US GDP reaches about $34.1 trillion in 2030. Unemployment stays inside historical ranges. The share of income going to workers rises slightly, to 59.4%.

Substantial scenario. AI becomes capable of doing about half of all knowledge work by 2030, most of it autonomously, but is not adopted for all of it. GDP reaches roughly $36.3 trillion. Unemployment rises to about 5%. Knowledge-worker wages are essentially flat while other workers see gains. Labour's share of income falls to 56.1%.

Extreme scenario. AI becomes more productive than humans at most knowledge tasks and is adopted rapidly. GDP reaches about $44.4 trillion — a very large economy. But unemployment spikes to historic recessionary levels, knowledge-worker wages fall more than 10%, and labour's share of income drops to 45.2% while capital's rises to 54.8%.

Anthropic itself describes the model as "a stark simplification of a complex reality" that excludes policy responses, business cycles, and financial disruption.

Goldman Sachs on the labour market

Goldman Sachs research, published in March 2026, estimated that AI could automate tasks representing about 25% of US work hours, with roughly 300 million jobs globally exposed to some degree of automation.

Its base case — assuming a roughly ten-year adoption timeline — is that 6% to 7% of workers are displaced, which would raise the unemployment rate by about 0.6 percentage points. Goldman's economists are explicit that a more front-loaded adoption path would produce much larger effects, and that the uncertainty is genuine.

Notably, the same research points in the other direction too: roughly 500,000 net new jobs are projected to be needed in infrastructure sectors by 2030, largely for power generation and data centre construction, and data-centre-related construction employment has already risen by about 216,000 since 2022.

What the Federal Reserve says

The July 2026 FOMC minutes are unusually candid. Among the risks participants discussed were the possibility that elevated AI-related equity valuations could reprice abruptly and tighten financial conditions, and that the net employment effects of AI remain significantly uncertain.

When the central bank writes down "we do not know" in an official document, that is the most honest number in the whole debate.

What "take over" would and would not mean

The phrase does a lot of work and most of it is unhelpful. Three distinct claims usually get bundled together:

Claim one: AI does a large share of economic work. This is what the scenarios above model, and versions of it are entirely plausible within five years.

Claim two: AI causes mass unemployment. This depends on adoption speed, on whether new work is created as fast as old work is automated, and on policy. Historically, technology has displaced tasks and created jobs, but the transition periods have been genuinely painful and unevenly distributed.

Claim three: AI systems act autonomously against human interests. This is a different category of concern with a different literature, and it is not what economic models address. Treating it as the same claim as the first two makes all three harder to think about clearly.

For financial planning purposes, the first two are the relevant ones.

The uncomfortable common thread

Look at all three of Anthropic's scenarios and notice what happens to the split between labour and capital.

In every scenario, capital's share of income rises. In the extreme case it rises enormously.

That is the single most financially actionable fact in the entire body of research, and it has an uncomfortable implication: the more disruptive AI turns out to be for workers, the more of the economy's income flows to people who own assets.

This is not a moral argument. It is an arithmetic one, and it has a plain consequence for an individual: owning a share of productive capital is a hedge against the value of your labour falling. If you only earn and never own, you are exposed to one side of that shift and not the other.

What a sensible investor actually does

1. Do not pick a scenario. You cannot know which one arrives, and neither can the people writing the forecasts — that is why they published three. Portfolios built on a single macroeconomic forecast fail when the forecast is wrong, which is most of the time.

2. Own the whole market rather than the story. In every scenario where AI creates enormous value, some companies capture it. Almost nobody correctly identifies which ones in advance. A total market index fund owns the winners automatically, at whatever weight they grow to, without requiring you to be right.

3. Recognise your job is the concentrated position. Your human capital — the present value of your future earnings — is usually your largest asset, and it is completely undiversified: one occupation, one industry, one set of skills. If that work is knowledge work, the scenarios above are describing a risk to your biggest holding. The responses are unglamorous: a larger emergency fund, a higher savings rate while earnings are good, and continued investment in skills that are complementary to these systems rather than substitutable by them.

4. Be sceptical of the concentration you already have. The market has already priced a lot of AI optimism. A handful of companies now represent an unusually large share of major indices, which means a broad index fund is a less diversified bet on this theme than it looks. That is worth knowing — not necessarily acting on, but knowing.

5. Keep the time horizon honest. 2030 is under four years away. Even in the substantial scenario, the world in 2030 is recognisably this world with a materially larger economy. Most of the wealth you will have in 2050 depends far more on your savings rate and your ability to leave a diversified portfolio alone than on correctly forecasting the next four years.

The honest conclusion

The serious forecasts do not say AI will "take over by 2030." They say the range of plausible outcomes has widened dramatically, that the upside case involves a much larger economy, that the downside case involves substantial labour market disruption, and that nobody — including the companies building the technology and the central bank regulating the economy — can currently distinguish between them.

Widened uncertainty is not a reason to make a dramatic bet. It is a reason to build a plan that does not require you to have made the right one.

Frequently asked questions