AI and Iran will speed up the energy transition

Iran and China are simultaneously driving a major energy transition

Control of the Strait Hormuz remains one of the biggest hurdles in US-Iran talks. (Reuters Photo: Majid Asgaripour)

Two things happened last week that give us a much clearer picture of what the world is going to be like from now on.

First, Iran asserted its control of the Strait of Hormuz by turning back two ships that didn't take the correct route and second, a Chinese company released an AI model that is as good as ChatGPT but a fraction of the price.

They don't seem linked, but they are. The Iran war and now Iran's control of the door to the Persian Gulf has changed global shipping forever and become a powerful accelerator for renewable energy and electrification, while the coming collapse in token prices — the currency of AI — driven by China will lead to faster and more widespread adoption of AI, and therefore much more demand for electricity.

China's AI start-up Zhipu's GLM 5.2 artificial intelligence model landed last week. (Reuters: Laurie Chen)

Iran, and probably Oman, will use the fact that the Strait of Hormuz lies within their maritime sovereign border to both exert geopolitical leverage and make money.

That means shipping large volumes of oil, or anything at all for that matter, is going become more difficult and more expensive.

That is especially so if other countries that have a narrow, busy strait within their territorial waters do the same, such as Malaysia, Singapore and Indonesia on both sides of the Strait of Malacca, and Eritrea, Djibouti and Yemen on both sides of the Bab el-Mandeb Strait at the entrance to the Red Sea.

Meanwhile in AI, China is about to do what China always does when it moves into an industry: prices fall, profits shrink and usage goes up.

For example, the Iran war isn't the only reason EV sales are surging in Australia this year — it's also because there are now about 30 Chinese EV models on the market from 12 manufacturers, with prices ranging from $23,000 to $130,000, and all of them half the price of similar European models.

The Iran war isn't the only reason EV sales are surging in Australia this year. (ABC News: Lachlan Bennett)

Americans need to rethink their business models

Companies have already been getting worried about how much they're spending on AI and are watching budgets and prices closely, so the new Chinese models are planting in some fertile soil.

In other words, unless governments decide to block Chinese AI for reasons of sovereign security, which is possible, and is already happening to some extent, China is likely to accelerate the uptake of AI and at best force the Americans to rethink their business models. At worst, it will destroy them.

The initial response to the recent release of Zhipu GLM-5.2 by a Chinese company known as Z.ai (presumably a play of Elon Musk's xAI) has been concern that it will burst the AI bubble on the share market.

GLM-5.2 is an "open‑weight model" that is now close enough to Claude and ChatGPT in performance that the gap is measured in single‑digit percentage points rather than generations.

An "open‑weight model" is one where the trained weights (that is the big "tensor", or digital files, used to store machine learning data and encode the model's parameters) are publicly released for download and local use.

Basically, the MIT licence lets you download, run, fine‑tune and use the model, but doesn't guarantee transparency about how it was trained or with what data, as it would if it was "open source".

For GLM‑5.2 specifically, "open‑weight" means that any developer or enterprise can pull the files from repositories like Hugging Face or GitHub, run them on their own infrastructure and agents or via third‑party providers, and fine‑tune or train it some more.

It won't necessarily destroy the business models of Anthropic, OpenAI and other large language model developers, but it certainly undermines them and will force them to adapt, probably by going upmarket.

In some ways it's like what happened with web browsers 20 to 30 years ago, leading to the so-called "browser wars". There's a good chance we'll have an AI war like that one, except this time involving Chinese firms vs American ones.

The analogy is apt, so it's worth recalling what happened with internet browsers.

The first browser war was between Netscape and Microsoft and started when Netscape Navigator launched in 1994. It was a huge leap in using the internet, and for 12 months it was the only game in town.

Microsoft panicked and built Internet Explorer, matching Netscape's features line-by-line. Then they bundled Internet Explorer for free into Windows 95 and 98, and overnight, no one had to download and pay for a browser because it was already integrated into the operating system they used. Microsoft famously called this "cutting off Netscape's air supply", which worked.

Ten years later, it was Google's turn to panic. Google built Chrome in 2008, sparking what became known as the second browser war, not because it wanted to be in the browser software business, but because its entire (by then) multi-billion-dollar business relied on internet search, via other peoples' browsers — mainly Microsoft Explorer, Firefox and, by then, Apple's Safari.

They realised that if Microsoft (via Internet Explorer) or Apple (via Safari) decided to change the default search engine to Bing or Yahoo, Google could have been wiped out overnight. So, Chrome was a defensive wall built to secure their gateway to the internet. Instead, Chrome virtually wiped out the other browsers because Google had the distribution, with most of the world's computers having Google as the home page.

In essence, the browser wars were about the importance of distribution over technology, and that still applies.

Maybe this time Google's and Meta's distribution isn't as much of a thing, because low barriers to entry have made distribution ubiquitous. Also, China itself is such a big part of the global market, and also has TikTok and WeChat.

As the implications of GLM-5.2 rippled through the market this month, share prices did start falling, although things were already looking wobbly after the shooting star that was SpaceX briefly lit up the sky.

Last week the US Nasdaq index fell nearly 5 per cent and SpaceX — the purest AI company that's publicly listed — fell 20 per cent (Anthropic and OpenAI are due to list this year and Alphabet, owner of Google Gemini, is a lot more than AI).

Of course, SpaceX's price had been tumbling since a couple of days after it listed, simply because it loses money and a A$4 trillion valuation was too ridiculous for even the most optimistic boosters of AI and Musk.

But while it will be a big deal if the bubble bursts and the stock market crashes, since these things have a habit of causing recessions, it's not the main thing about AI.

We don't yet know whether it will cause mass unemployment or the extinction of humanity, but one thing we know for sure is that it is causing a structural increase in the demand for electricity.

The combination of this with a new era of more difficulty and expense in transporting fossil fuels around the world is likely to accelerate the energy transition in a way that 30 years of UN-sponsored global climate change conferences has failed to do.

Alan Kohler is a finance presenter and columnist on ABC News. He hosts the podcast That’s Business with Alan Kohler in the ABC Business Daily feed on Friday. He also writes for Intelligent Investor.