International | Sep 18 2025
This story features AGL ENERGY LIMITED, and other companies.
For more info SHARE ANALYSIS: AGL
The company is included in ASX100, ASX200, ASX300 and ALL-ORDS
OpenAI represents a suite of opportunities and risks for big Tech and investors as AI adoption is changing the world as we know it.
-From startup to US$300bn giant, OpenAI's ChatGPT rises to 700m users
-Google’s Gemini gains momentum as competition heats up
-Hyperscaler capex and Oracle’s US$30bn bet on OpenAI
-Market Bubbles, infrastructure strains, and AI’s global footprint
By Danielle Ecuyer
What a difference three years makes
Cast your mind back to November 2022; markets had just started to recover from the post-covid interest rate hiking cycle as central banks wrestled to pop the inflation genie back in the bottle.
It was also the month OpenAI launched ChatGPT, the now ubiquitous ‘AI’ platform which recently locked in another US$8.3bn of private equity and venture capital, reported as one tranche of a US$40bn funding round announced earlier in 2025.
Since then, the US stock market has risen by over US$21trn driven largely by the Magnificent Ten; the world’s largest technology companies riding the AI investment boom, with a plethora of smaller companies benefitting from the use of software, hardware and energy solutions as the infrastructure platform is built out to support AI.
The emerging AI behemoth, the puppet or the puppet master?
OpenAI was valued at US$300bn as at the beginning of August and ChatGPT has 700m-plus weekly average users, up from 200m a year ago, representing around 10% of the world’s population, achieved in under three years.
Already, this marks one of the most significant growth ramp ups in the technology sector. In a recent podcast ChatGPT’s head said, “we’ll hit a billion users soon”.
In the words of UBS, OpenAI is “one of the most consequential AI firms”, arguably one of the most consequential companies in the history of technology. It is the single largest customer of Nvidia GPUs, which now has a US$4trn-plus market capitalisation, with a growth tailwind rippling through the world’s largest technology companies.
Oracle’s latest quarterly report was a standout case in point. Despite ho-hum earnings, investors jumped on the company’s outlook statement including US$30bn of revenues coming from one customer in FY28, which has been assumed is OpenAI.
The stock price rallied 40% post the update, taking Oracle’s valuation over the US$800bn mark.
The growth of OpenAI and ChatGPT is inextricably linked to the fortunes and potential disruption of even America’s largest companies.
In a recent in-depth update on the AI industry, UBS takes a deep dive into OpenAI’s growth prospects and the likely impacts on tech stalwarts and behemoths including Microsoft, Oracle, Google, and the software sector at large.
Taking a step back, let’s wrap some financial metrics around the company to contextualise OpenAI’s current and potential impacts.
Introducing: OpenAI, the centre of today’s tech boom
Given OpenAI is a private company, analysts rely on credible media sources for latest information.
Estimated revenue for 2025 of US$13bn is quickly fading into the rear-view mirror with the New York Times noting OpenAI had recently passed US$12bn in run-rate revenue and could reach US$20bn in run-rate revenue by the end of 2025.
This infers a 2025 full year revenue run rate of US$15bn, compared to estimated revenue of US$3.4bn in 2024. It equates to a 275% growth rate. The company generated US$1bn in revenue in 2023.
Is it too far-fetched to project OpenAI can grow its revenue to US$129bn in 2029 which would be around the size of Amazon’s cloud service AWS currently?
Not according to UBS, however, it would necessitate a compound average growth rate per annum of 70% which has never been witnessed at this scale.
OpenAI pointed to a rise of 60% growth in two months to over five million ChatGPT business paying users in August from three million in June. The company has also recently announced measures to expand into the enterprise segment with features like Operator agent and Deep Research tool.
ChatGPT can now link with aspects of Google Workspace including Gmail to apply the service to acting on emails and documents.
For observers, this has been highlighted as a push into taking on the large employees’ productivity market, dominated by both Google Workspace and Microsoft M365/Office 365.
Compared to an estimated 35m ChatGPT users, five million business users represent some 15% of the total which would align with management’s statement that circa 25% of revenues came from enterprises.
While you might be thinking this is all conjecture and blue-sky forecasts, the rise of OpenAI is almost single handedly supporting the likes of Microsoft, Oracle, Nvidia, Google Cloud and more recently listed companies, like Coreweave, as well as software firms that serve it, like Datadog, Snowflake and Databricks.
If as the media reports, OpenAI is targeting revenue of US$86bn in FY28, a US$30bn spend with Oracle would represent 35% of revenue.
Industry sources have suggested OpenAI is investing 50% of revenues or around US$45bn on compute over the next three years. A shift from 100% use of Microsoft to 67% use of Oracle would have a major impact on Microsoft Azure and Oracle Cloud Infrastructure growth, and probably on their respective share prices too.
Success is increasingly linked to the success of all the ancillary technology companies that are benefitting from OpenAI’s growth. Equally, it presents a risk in terms of both disruption and failure to achieve expected growth rates.
According to UBS, market opportunities for AI products remain both “attractive and still under-penetrated”.
Google announced on its second quarter earnings call it had 450m monthly average users on its own AI applications (Gemini) and over 100m users for its AI mode feature in its proprietary search toolbar.
Media reports suggest around 35m xAI Grok monthly average users and circa 30m for Anthropic’s Claude users, although the latter’s user base is more slanted to enterprise operations such as coding.
OpenAI is considered unique given the rarity of a private company achieving such scale and influence including the flow through to large publicly listed companies.
Latest data highlight the changing impacts of AI
According to a working paper from National Bureau of Economic Research, analysing 1.5m ChatGPT users’ conversations from June 2024 to July 2025, 70% of usage is personal, not professional.
The percentage of users with normally feminine names rose to 52% in July 2025 from 37% in January 2024.
Historically, technology reporter and consultant Shelly Palmer explains, when women represent most users for any technology platform, there is “true mass market adoption”.
Messages “seeking information” now represent 24% of usage, up from 14% on the prior year. Palmer notes the research emphasises this result “appears to be a very close substitute for web search”.
This in turn poses a genuine threat to Google’s position and search popularity.
In contrast, the latest update at May-end showed Google’s Gemini was experiencing around 6m installs per week on Android while ChatGPT’s downloads had dropped below 3m per week.
MarketWatch reported the weekend past (Sept 13-14) Gemini’s Nano Banana which launched August 26 had already become the most popular free iPhone app via Gemini’s artificial assistant and Gemini has gained 23m users and edited 500m images.
While Google may have an advantage in using software and AI to enhance digital images due to a decade of experience, the example underscores the rapid rate of change and potential disruption AI is creating.
Confirming the trend that AI adoption is following mobile phone penetration patterns, Palmer points out growth in ChatGPT usage has been highest in the lowest income countries at more than four times the rate of the highest income countries.
Palmer asserts the data should inform assumptions about AI adoption and how users access ChatGPT. The three main usage categories from the data are information seeking at 24%, writing at 24% and practical guidance at 29%, representing 77% of all conversation use. Technical help including programming declined to 5% from 12% over the year.
Consumer adoption is expected to increase workplace adoption.
Historical search and discovery methods are increasingly at risk of being bypassed.
The twists and potential turns of OpenAI’s impact
Summarising UBS’s research effort, the broker proposes Microsoft and Azure have benefited from partnership with OpenAI, but as the latter shifts its GPU capacity to Oracle, Coreweave and Google, aspects of the partnership could well be re-negotiated.
ChatGPT is also potentially an alternative to Microsoft’s Copilot at some stage.
Nvidia generates data centre revenues of US$165bn with a large scale and breadth of customers for its GPUs. However, OpenAI is one of its largest single customers and thus its success remains a tailwind for Nvidia.
In terms of software companies, there has been considerable market concern over recent months that those exposed to infrastructure, data and cyber security end markets will benefit from the rise of OpenAI and be less disrupted, a la the potential impact on Copilot on software-as-a-service and application software providers.
Across UBS’ channel checks, there is little evidence of “displacement” anecdotes with no enterprise IT executive desks pointing to AI applications displacing incumbent SaaS applications.
The market view that AI models are eating software is seen as overhyped. While sensitive to these concerns, UBS believes the market is being overly cautious and discounting relevant stocks too much.
What is underpinning all this AI compute?
As described by McKinsey & Company, data centres are the “foundational infrastructure that fuels the many digital services” including not only the transition to the cloud, but compute for AI.
McKinsey’s research estimates US$6.7trn of capital expenditure will be invested globally in data centre infrastructure through to 2030 as the demand for capacity is projected to triple by then.
Around 70% of demand will come from hyperscalers (Amazon, Google, Oracle, Microsoft and Meta) with GenAI demand forecast at around 40% growth through to 2030.
More than US$4trn of investments will go to computing hardware with the balance into real estate and power infrastructure; with more than 40% destined for the US.
Morgan Stanley estimated as at the end of July the cost to bring more data centres online could reach US$2.9trn between 2025-2028.
The latest quarterly US reporting season revealed substantial growth in AI infrastructure spending across the hyperscalers. Bank of America places 2025 capex at US$414bn, up 44% on the prior year.
The scale of the financial metrics has become eye-wateringly large for investors and has inevitably prompted discussions around “bubbles”, capacity constraints, and challenges across multiple factors, including financing, energy and water.
In terms of bubbles, The Economist’s recent update on the topic highlighted the current AI spending, while “worrying” remains relatively conservative within an historical context.
US AI companies have invested some 3-4% of GDP in the last four years compared to investment in British railways during the 1940s representing around 15-20% of GDP. What is probably more concerning is the expected rate of technological change in infrastructure, the author postures.
Will Nvidia’s leading edge GPUs become outdated and clunky in a few years?
Some investors would respond that is why this company has a never-before-seen new product cadence of updates every 12-months.The shelf life for your average US tech firm’s assets is believed to be no farther than nine years in contrast to telecom assets of 15 years in the 1990s.
Questioned whether AI is a bubble, Viktor Shvets, Global Strategist at Macquarie Capital who was a keynote speaker at Livewire Markets’ recent conference in Sydney, made the following comments (should be read within the greater context of his presentation (https://youtu.be/XtEqBCHmtMY?si=KHwbqWYCLB32of_u))
“… a lot of people are worried about the bubble of AI, no question that any general-purpose technology, and AI is a general purpose technology, requires a bubble to progress, and this will always be the case.
“The only difference now is we have never had access to abundant capital and when something is abundant it doesn’t have a price, so that means some of the extremes will never come to pass because there is too much capital chasing incremental returns.
“The second thing is how quickly you can create companies and destroy companies is unique. Those two things imply you might have rolling bubbles rather than one bubble collapsing, bringing down the market with an expensive period of recuperation, there is instead another bubble almost as quickly.
“For example, if Meta loses -30% of market capitalisation and Palantir triples, there is almost no effect, you just don’t want to be in Meta.”
Shvets’ proposition does not necessarily align with mainstream experts or narratives but at the very least encapsulates the major megatrends shaping markets, disruption, technological change alongside a heavily financialised global system.
Risks and capital funding requirements
Funding these investments is equally pertinent for investors, in terms of where will the available capital come from and if it fails to be successfully monetised, who will be burdened with the impairments and losses?
Morgan Stanley estimates just under 50% of the forecast US$2.9trn in data centre investment from 2025-2028 will be funded via cash generated from hyperscalers.
As previously discussed. much of OpenAI’s investment spending will be internally generated from expected revenues.
While some commentators are drawing parallels to a pending credit bubble in data centres like the mid-to-late 1990s which led to the subsequent Dotcom bust, Morgan Stanley proposes the diversity of the capital pool is considerably deeper than thirty years ago, when a lot was sitting on corporate balance sheets.
Equally, the quality and relatively secular not cyclical cash flow generation of hyperscalers makes them less sensitive to macro conditions.
Return on investment from AI spend should be positive, generating US$50bn in revenues in 2025, forecast to exceed US$1trn by 2028.
Other sources of capital are likely to be insurance companies, pension savings capital pools, sovereign wealth funds, family offices. Meta’s US$29bn data centre expansion in Louisiana is being funded by Pimco and private credit firm Blue Owl.
The Economist highlights American banks are not financing the AI boom with most exposure through non-bank lenders.
Implications for energy and water usage around data centres
Energy utilities have equally been caught out by the rise in power demand from the growth in data centres.
Wood Mackenzie highlights the mismatch between technology companies’ demand visibility for three to five years out compared to energy investors’ longer investment time frame of 30-years.
As of June 2025, the consultants were tracking 143GW of proposed data centres in the US, up from 50GW in the prior year.
The scale of growth poses challenges not only for energy producers but also the regulatory environment across transmission, interconnection and approval for expansion plans.
US utilities have committed to supply 64GW of new data centre capacity which equates to a 12% increase in current electricity demand.
The Financial Times noted a report from Jefferies that utility capex is expected to reach US$212bn in 2025, a rise of 22.3% on 2024 and up 129% compared to a decade ago. Investment is anticipated to lift to US$228.1bn in 2027.
The real challenge will be managing the growth in demand and associated investment spending against how much of the costs are passed onto households and small businesses.
Consulting group ICF has estimated US electricity demand will grow 25% by 2030 and 78% by 2050 from 2023 levels, while residential prices are flagged to increase between 15% and 40%.
Business Insider projects US data centres could consume more electricity than Poland with a population of 36.6m in 2023.
Morgan Stanley recently indicated in its AGL Energy ((AGL)) update that year to date NSW night demand for electricity is up 77MW on the prior year, which is a proxy for data centre demand.
Australia is undergoing its own major data centres build out as yet again featured in August result releases from the likes of NextDC ((NXT)), Infratil ((IFT)), Goodman Group ((GMG)) and Macquarie Technology ((MAQ)), to name a few examples.
See also FNArena’s dedicated GenAI section (https://fnarena.com/index.php/tag/gen-ai/)
Turning to water, Morgan Stanley also noted usage in data centres for cooling and electricity generation is likely to see consumption rise eleven-fold from 2024 to 1,068bn litres by 2028 in the US.
The analysis also points to a rising water footprint for semiconductor manufacturing.
While significant, AI’s water consumption is expected to remain “modest” compared to traditional water withdrawals across major sectors such as irrigation, domestic use and livestock.
Conclusion, as written by ChatGPT
OpenAI has become the epicentre of the AI investment cycle, driving extraordinary growth in hyperscaler capex and reshaping valuations across Big Tech.
With ChatGPT’s user base surpassing 700m weekly actives and revenues on track to more than double in 2025, the company is now valued at US$300bn.
Its outsized spend on compute is directly benefitting Nvidia, Microsoft, Oracle and other infrastructure players, while raising questions about sustainability, capital intensity, and potential bubbles in AI.
Rival platforms such as Google Gemini are gaining momentum, underscoring how quickly competitive dynamics in the sector can shift.
Find out why FNArena subscribers like the service so much: “Your Feedback (Thank You)” – Warning this story contains unashamedly positive feedback on the service provided.
FNArena is proud about its track record and past achievements: Ten Years On
Click to view our Glossary of Financial Terms
CHARTS
For more info SHARE ANALYSIS: AGL - AGL ENERGY LIMITED
For more info SHARE ANALYSIS: GMG - GOODMAN GROUP
For more info SHARE ANALYSIS: IFT - INFRATIL LIMITED
For more info SHARE ANALYSIS: MAQ - MACQUARIE TECHNOLOGY GROUP LIMITED
For more info SHARE ANALYSIS: NXT - NEXTDC LIMITED