An Australian bank’s research desk has just put a bigger number on one of China’s most closely watched artificial intelligence companies. Macquarie has raised its revenue target for Zhipu AI, the Beijing developer behind the GLM family of large language models, on the strength of demand for its coding tools, according to a note reported by Investing.com Australia.
The upgrade is a small item in the daily churn of equity research, but it sits at an interesting crossroads. It ties an Australian financial institution to the Chinese AI race, and it flags the same trend that has reshaped budgets in software teams from Sydney to Silicon Valley: coding has become the killer app for large language models.
Who is Zhipu, and why the coding angle matters
Zhipu AI, which markets some of its products internationally under the Z.ai brand, is one of a cluster of Chinese startups often grouped together as the country’s “AI tigers.” Spun out of Tsinghua University research, it has built the GLM series of models and has been moving toward a public listing, making analyst coverage of its revenue trajectory more than academic.
The reason Macquarie’s analysts have singled out coding is straightforward. Of all the things generative AI can do, writing, completing and debugging code has proven the fastest to translate into paying customers. Developers will happily pay for a tool that shaves hours off their week, the output is easy to check, and the productivity gain is measurable in a way that, say, a chatbot’s tone is not. That has turned coding assistants into the most commercially dependable slice of the AI market, and it is where much of the competitive heat now sits.
Zhipu has leaned into that. Its GLM coding models have been positioned as lower-cost rivals to the Western incumbents, and the company has courted developers with open weights and aggressive pricing. For a research desk trying to forecast revenue, evidence that those tools are being adopted at scale is exactly the kind of signal that justifies a higher target.
Two ways to read the upgrade
The bullish reading is that Zhipu has found genuine product-market fit in a segment that generates recurring, sticky revenue. If developers build workflows around a given model, switching costs rise, and a company with a credible coding product can compound usage quickly. On that view, Macquarie is simply catching up to a business that is growing faster than earlier models assumed.
The sceptical reading is worth holding alongside it. China’s AI sector is crowded, margins on inference are thin, and the coding category is fiercely price-competitive, with Chinese players undercutting one another as much as they undercut the Americans. A revenue target is not a profit target. Rapid top-line growth funded by cut-price tokens can flatter a company that is still burning cash to buy market share. There is also the geopolitical overhang: US export controls on advanced chips constrain how cheaply Chinese labs can train and serve frontier models, and Washington has floated restrictions on Chinese AI software reaching Western users. Any forecast has to be read through that lens.
Neither view is settled, which is precisely why the analyst call is interesting rather than routine. It is a considered bet, from an Australian institution, that the coding wave has further to run and that Zhipu is a real beneficiary of it.
The Australian stakes
For Australian readers, the story lands on two levels. The first is the Macquarie name itself. Macquarie is one of the country’s marquee financial exporters, and the fact that its research is shaping the narrative around a Chinese AI champion is a reminder that Australian capital and analysis are woven into the global AI trade, even when no Australian company is in the frame.
The second, and more consequential, level is what the coding boom means for local software teams. Australian developers, whether at the big banks, at fast-growing startups or inside government agencies, are already heavy users of AI coding assistants. The dominant tools in the Australian market skew Western, with Anthropic’s Claude and offerings from OpenAI, Microsoft’s GitHub Copilot and Google prominent in enterprise contracts. A credible, cheaper Chinese alternative changes the calculus, at least on price. It also sharpens a question Australian chief information officers are increasingly asked to answer: where does your code, and the context you feed the model, actually go?
That question is not hypothetical here. Australia has spent much of the past year debating AI sovereignty, data residency and the risks of routing sensitive workloads through models hosted offshore. Local firms including Xero have signed on with Anthropic, Commonwealth Bank has deepened its Microsoft AI partnership, and the federal government has been sketching out national frameworks and a proposed Office of AI. In that environment, a low-cost Chinese coding model is unlikely to walk into a regulated Australian bank or a Commonwealth department, whatever its benchmark scores. For a startup or an offshore-friendly software shop watching its cloud bill, though, the pricing pressure is real, and it filters back into what the incumbents can charge everyone.
There is a competitive-dynamics point too. Every dollar Zhipu and its Chinese peers strip out of the cost of AI coding is a dollar of pricing power removed from the Western labs that Australian enterprises rely on. Cheaper models abroad tend to mean cheaper models at home, eventually, which is good news for the Australian businesses trying to justify their AI spend and awkward news for the vendors banking on premium margins.
What’s next
The near-term test is whether Zhipu’s revenue actually tracks toward the higher figure Macquarie now has pencilled in, and whether the company can convert developer enthusiasm into durable, profitable accounts rather than a churn of bargain hunters. A move toward a public listing would put far more disclosure on the table and let the market judge the growth story on hard numbers.
For Australia, the thread to watch is adoption. If Chinese coding models keep improving and keep getting cheaper, the pressure to at least trial them will grow, and so will the governance questions that come with them. Expect procurement teams, regulators and the vendors themselves to spend the rest of the year working out where the line sits between cost and control. Macquarie’s upgraded target is one small data point that the wave is still building.
Sources: Investing.com Australia.
