Australia’s Data Centre Boom Marches On

Array
(
    [0] => Array
        (
            [0] => ((MQG))
            [1] => ((CBA))
            [2] => ((IFT))
            [3] => ((NXT))
            [4] => ((MAQ))
            [5] => ((DGT))
        )

    [1] => Array
        (
            [0] => MQG
            [1] => CBA
            [2] => IFT
            [3] => NXT
            [4] => MAQ
            [5] => DGT
        )

)

This story features MACQUARIE GROUP LIMITED, and other companies.
For more info SHARE ANALYSIS: MQG
The company is included in ASX20, ASX50, ASX100, ASX200, ASX300 and ALL-ORDS

CommBank's transition to the cloud is symbolic of the systemic shifts across enterprises as they position in the cloud for the AI era.

-AI ‘bubble’ fears and why today’s market looks different from 1999
-Cloud migration, CommBank, and the drivers of enterprise demand
-Power, regulation, and capacity: Australia’s growing data centre challenge
-Neoclouds and CoreWeave: new players reshaping AI infrastructure economics

By Danielle Ecuyer

AI bubble has become the prevailing market narrative

Before examining the composition of Australian data centre companies and recent updates, it is worth revisiting the narratives around AI, technology stocks, and valuations.

Arguments supporting a Dot-Com style crash are well cited in mainstream media, but at FNArena we draw on research from major financial institutions to frame the discussion.

Ed Yardeni of Yardeni Research recently posed the question: “Is there a Bubble in Bubble Fears?”

Google Search trends show queries for “AI bubble” spiking from virtually non-existent in mid-September to a peak by October 2. Yardeni suggested today’s technology-related stocks “have less air than the one during 1999”.

Currently, the S&P500’s IT and Communication Services sectors account for 44.9% of market capitalisation and 37.4% of forward earnings, versus 40.7% and 23.8% in the 1999–2000 bubble. Concentration is high, but earnings support is stronger.

Morgan Stanley strategist Michael Wilson argues the AI capital investment cycle that began in 2023 marks a “new era for equity investors”.

He believes payoffs from these mega-trend investments will come in time, with stocks already signaling “trust me—it’s not a question of if, but when”.

Wilson views April 2025 as the trough of the US rolling recession since 2022, coinciding with a slowdown in AI spending. Earnings and valuations bottomed at the same time.

He believes consensus earnings forecasts for 2026 remain too low.

His contrarian view is that a higher inflationary era will underpin equity valuations via lower equity risk premiums. While US stocks appear expensive on price-to-earnings ratios, higher nominal GDP and earnings growth with inflation provide valuation support.

Wilson also highlights median large-cap free cash flow yields are nearly three times higher than in 2000, while margins and profitability appear much stronger.

“Free cash flow generation, operational efficiency and strong profitability are all characteristics of a higher quality index than we saw in the late 1990s.”

Equities and gold remain key inflation hedges.

Australia’s narrower AI exposure, so far

Australian investors face a more limited universe of AI-linked stocks, particularly data centres and companies such as Macquarie Group ((MQG)) that invest in them.

Jeff Bezos recently highlighted the breadth of AI’s impact, suggesting it will touch “every company in the world” and boost both quality and productivity.

Locally, more companies are reporting how AI functionality is being embedded into their operations. Morgan Stanley highlights Anthropic’s Economic Index, which tracks Claude usage across US states and industries. As of September 2025, directive automation rose from 27% to 39% of conversations in nine months, and to 77% for enterprise customers.

Extrapolating Anthropic’s work, Morgan Stanley estimates AI adoption could add US$920bn in US pre-tax profits for the S&P500, with a long-term value of US$13–16trn.

Analysts draw two conclusions: the pace of improvement will be non-linear and use cases will involve both augmentation (AI complementing human tasks) and replacement (automation and robotics), with the latter expected to be more of a mid-2030s development.

While the AI story still has a long way to unfold, the immediate race is to build the infrastructure —data centres— that will underpin the computational power needed to support large language models and the next phase of technological evolution.

CommBank exemplifies major structural trends

As evidenced this week, the transition of enterprises to the cloud is still progressing. CommBank ((CBA)) is Australia’s first bank and one of the largest global banking transitions to the cloud, managed by Amazon’s AWS.

The transformation was highlighted by the bank’s group executive for technology as “a cornerstone for our enterprise transformation” and “innovation is faster on the cloud,” he told the AFR.

The announcement follows a partnership with OpenAI in August, while the bank’s data are centralised in a “data lake” run on a Snowflake system in AWS.

CommBank demonstrates two major forces driving demand for data centre capacity: the staged, ongoing shift of enterprises to the cloud, and the sharp rise in demand for AI-centric infrastructure, which requires immediate storage and compute rather than phased rollout.

Implementing large language models is not staged; capacity is needed upfront, often referred to as the “everything, everywhere, all at once” model.

How the Australian data centre trend is evolving

Data centre operators are balancing both demands, building more infrastructure to meet rising requirements, while also managing customer-specific needs that directly affect earnings.

Cloud migrations require heavy upfront capex, with earnings booked over time, whereas AI workloads accelerate income as contracts are realised.

At a macro level, Macquarie points to regulatory changes at both federal and state levels supporting a commitment to the OpenAI Stargate Global deal and to government data centre capacity growth.

A case in point is Canberra Data Centres (CDC), 49.75% owned by Infratil ((IFT)), as noted in a recent company update, see https://fnarena.com/index.php/2025/10/03/infratils-digital-decarbonisation-drive/

“This business is gaining from the Federal Government’s increased backing of digital infrastructure and is progressing a 200MW high-density data centre project located south-east of Perth’s CBD.

The facility is strategically positioned to support AI and security workloads linked to AUKUS defence cooperation, leveraging its secure infrastructure and close proximity to the Henderson Naval Base.” 

The analyst estimates 200MW to 750MW of market demand. Rising AI-driven capacity reflects innovation and new data centre reference architectures. Hyperscalers like AWS rely on operators for capacity that determines “time to market” for AI tools and products.

Infrastructure design is also evolving, shifting from air to liquid cooling and higher rack densities. This both improves hyperscale returns and boosts data centre profitability, as the same MW of capacity requires less floor space. Macquarie considers AI capacity commoditised in the medium term, but industry feedback suggests near-term contract pricing is being dictated by hyperscalers’ urgency.

E&P remains positive on Australian and regional data centre markets, a stance it has held for over a decade. Australia is now regarded as a major data centre market with Sydney and Melbourne both Tier 1 hubs.

While some assumed hyperscaler spend was receding in Australia when Microsoft’s OpenAI-related pause slowed global buildouts and pushed contracts forward, both E&P and Macquarie expect capacity growth to resume strongly, forecasting 5GW of new build by 2027.

Citi recently upgraded its hyperscaler capex forecasts through to 2029 by 20% to US$2.8trn, with 2026 forecasts raised by 17% to US$460bn, providing a strong tailwind for Australian operators. Globally, capacity is forecast to expand at a 57% CAGR from 2GW to 30GW. Citi believes Australia will grow faster than APAC due to geopolitical positioning and power availability.

E&P, however, cautions power constraints will emerge. While currently well supplied, the surge in data centre demand could strain the system.

The Australian Energy Market Operator forecasts data centre demand could exceed 5.2GW in an accelerated case, or 2GW in a slower scenario.

By 2055, data centres could use up to 18% of total power. However the decommissioning of over -11GW of nameplate capacity over the next 10 years according to the Federal Government’s accelerated transition case would already result in data centre energy use around 16% of total baseload power in the grid by 2035.

In the US, data centres consumed 4.4% of total power in 2023 and the percentage could rise to 12.5% by 2028. In Australia, over 1GW of new Sydney capacity is planned, including CDC’s 700MW Marsden Park site and NextDC’s ((NXT)) 400MW S7, while Melbourne has over 1GW in the pipeline across CDC, AirTrunk and Stack.

This is in addition to more than 2GW already built or under construction. E&P regards the 2GW national forecast out to 2054 as very conservative, noting seven operators —NextDC, Equinix, Global Switch, Digital Australia Investment Management, AirTrunk, AWS and CDC— carried about 500MW of workloads during 2023-2024.

Grid stability and transmission are already concerns. Substations in Macquarie Park have reached capacity for new customers, with a new substation only expected once four new data centres are connected in December 2028.

E&P sees Melbourne’s market about to “take off”, with growth unrelated to local population but linked to hyperscaler activity. AWS has already set up a machine learning hub there. Sydney remains in high demand as a global Tier 1 market, with Microsoft, AWS, Alphabet, Oracle, Meta and ByteDance operating at scale.

Despite some macro concerns, brokers remain positive. E&P rates NextDC Positive with a $28.66 valuation, Macquarie Technology ((MAQ)) at $112, and Infratil at $12.19. DigiCo REIT ((DGT)) is rated Neutral at $3.61.

Citi has upgraded Megaport to Buy with a $16.30 target, expecting it to benefit from both Alicloud’s flagged tenfold data centre expansion by 2032 and the growth in inference compute, projected to rise from 30% of workloads in 2026 to 50% by 2028.

Citi rates NextDC Buy with an $18.35 target. Macquarie rates NextDC Outperform with a $20.90 target, with Infratil and Macquarie Technology also rated Outperform at NZ$13.26 and $97.30 respectively.

DigiCo is rated Outperform with a $3.90 target.

The rise of neoclouds, no threat to data centres

Neoclouds have recently been in the spotlight with the Nasdaq IPO of CoreWeave in March, priced at US$40 per share and raising US$1.5bn. At the IPO, the market capitalisation stood at US$23bn.

The stock is currently trading at around US$134 with a market capitalisation of US$66bn. CoreWeave offers a cloud platform purpose-built for running AI workloads, including model training and inference.

Neoclouds provide a GPU-as-a-service model, described by DCByte as “ultra-low-latency networking and software stacks for large-scale AI workloads.”

While hyperscalers are still expected to dominate, neoclouds are offering faster access to GPUs and more flexible service models. Rather than replacing hyperscalers or data centres, they are capturing AI workloads that continue to be housed within data centres.

CoreWeave has attracted criticism around financial metrics, as it uses debt to fund the GPUs which are then contracted out.

The company requires high utilisation of its GPU infrastructure to spread fixed costs and underpin margin expansion. The company remains unprofitable and is reliant on realising its revenue backlog.

By way of example, DigiCo’s recent contract update showed customers customers signed span various segments from hyperscale, neocloud, enterprise, and government. 

Latest news from OpenAI over the last week, care of ChatGPT

OpenAI has been at the centre of major AI infrastructure and strategic news in the past week. The company signed a landmark partnership with AMD, agreeing to purchase its upcoming Instinct MI450 chips and secure up to 6GWs of compute capacity from 2026.

As part of the deal, OpenAI also received warrants for up to 160m AMD shares, potentially worth a 10% equity stake if performance milestones are achieved. The agreement reflects OpenAI’s push to diversify hardware suppliers beyond Nvidia, whose GPUs have long dominated AI training.

The company also sharpened its enterprise focus at its latest DevDay event, unveiling a ChatGPT “app store” that allows third-party developers to build, distribute, and monetise apps within the platform.

Partnerships with firms such as Spotify and Zillow highlight the breadth of integrations OpenAI is now pursuing, positioning ChatGPT as a gateway for consumer and enterprise AI adoption.

OpenAI’s valuation surged to about US$500bn after a secondary share sale in which current and former employees sold roughly US$6.6bn worth of stock. That jump cements OpenAI’s status as one of the most valuable private technology firms in the world.

Infrastructure remains a global story, with OpenAI expanding its “Stargate” program through new partnerships in South Korea, including deals with Samsung and SK Group to build data centre capacity and secure chip supplies. These agreements signal how OpenAI is aligning national industrial policy with its own infrastructure needs.

At the same time, OpenAI’s safety protocols are under renewed scrutiny. Academics reviewing the company’s Preparedness Framework argue it still permits deployment of high-risk models without enforceable safeguards, raising questions about whether governance is keeping pace with technological expansion.

Together, these developments highlight OpenAI’s dual track: racing ahead with compute supply deals, global partnerships, and platform expansion, while critics warn safety and regulatory oversight may lag behind its rapid scaling.

A question for everyone to ponder, has the world ever seen a start up moving at such speed with such aggressive expansion plans?

****

For more reading on related AI topis, refer to FNArena’s dedicated GenAI section at https://fnarena.com/index.php/tag/gen-ai/

****

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