Equity Mates Investing Podcast
Equity Mates Investing Podcast

Nov 19, 2026 · 37 min

AI infrastructure shifts investor focus beyond the chipmakers

Chips, power & who gets paid in the AI build-out with William Taylor

The AI build-out depends on power, cooling, construction and financing as much as computing hardware, creating both opportunities and risks for investors.

3 key takeaways
  1. 1Hyperscaler spending is increasingly supported by revenue, but debt and equity issuance leave the investment cycle’s durability unresolved.
  2. 2Australia could capture infrastructure investment through land, renewable energy, stable institutions, subsea cables and existing hyperscaler operations.
  3. 3Power, approvals and construction timelines are the central bottlenecks, while neocloud financing creates more contained risks than a system-wide threat.

Don't miss

Taylor breaks down how an AI data center depends on grid connections, transformers, electricity, cooling and construction before computing can run.

The brief

William Taylor of ETF Shares frames the AI infrastructure debate around two questions: whether spending has become a bubble and whether investors have already missed the opportunity.

His answer is more measured than either a boom narrative or a collapse call: companies are stronger than during the dot-com era, but earnings durability and pricing power remain untested.

The physical build-out reaches well beyond GPUs, spanning networking, transformers, power generation, cooling, construction and data-centre operators—areas the ETF targets instead of neoclouds.

Australia emerges as a potential beneficiary through available land, renewable energy, stable institutions, subsea cables and established hyperscaler operations, though approvals and grid capacity remain constraints.

The clearest investment lesson is portfolio discipline: treat infrastructure as a smaller growth allocation, diversify exposure and do not mistake every pullback for a collapse.

What was said on this episode

25 statements · 12 positive · 7 negative · 6 neutral

  1. Will Tayloron AI data centersNegative0:02

    An AI data center consumes power equivalent to 100,000 households.

    “An AI data center uses the same amount of power as 100,000 households.”

    Listen at 0:02

  2. Will Tayloron AI infrastructure marketsNeutral4:20

    AI infrastructure markets are currently in an execution phase.

    “I would say we're kind of in that execution phase.”

    Listen at 4:20

  3. Will Tayloron AI infrastructure investmentsPositive5:12

    The AI infrastructure pullback may offer investors a favorable buying opportunity.

    “It's a really good time to actually potentially buy it because the market has had this kind of pullback”

    Listen at 5:12

  4. Will Tayloron AustraliaPositive10:12

    Australia could benefit substantially from the AI infrastructure buildout.

    “I actually think Australia can do really well here.”

    Listen at 10:12

  5. Will Tayloron AI infrastructure valuationsPositive13:04

    AI infrastructure valuations are not sufficiently detached from fundamentals to constitute a bubble.

    “I wouldn't say it's drifted so much that we're now kind of in bubble territory.”

    Listen at 13:04

  6. Will Tayloron AnthropicPositive14:57

    Anthropic’s quarterly revenue increased approximately tenfold year over year.

    “If we look at Anthropic, for example, last September, their quarterly revenue was just under $2 billion. This year's one quarter, a September quarter, and it was up, I think it increased tenfold.”

    Listen at 14:57

  7. Will Tayloron AI modelsNegative15:23

    AI models becoming interchangeable would reduce providers’ pricing power.

    “does it become a commodity? Because when it does, and like a model is just a standard, doesn't matter if you go Gemini, Claude, OpenAI, if it's all giving you the same answer, then where does the pricing power go?”

    Listen at 15:23

  8. Will Tayloron AI infrastructurePositive15:44

    AI infrastructure will be built regardless of which model provider wins.

    “They get built regardless of whose model wins.”

    Listen at 15:44

  9. Will Tayloron Neocloud spendingNeutral17:01

    Neocloud spending is relatively small compared with total AI infrastructure spending.

    “If we look at the NeoCloud spending versus the overall picture, it's quite a small amount.”

    Listen at 17:01

  10. Will Tayloron Australian data-center projectsNegative20:03

    Australian proposed data-center projects require twice the grid’s peak demand.

    “the amount of power that they require is double the peak demand of the entire grid.”

    Listen at 20:03

  11. Will Tayloron AI data centersNegative20:51

    AI data centers consume roughly ten times more power than traditional data centers.

    “the power consumption is 10 times as much as like a traditional just data center.”

    Listen at 20:51

  12. Will Tayloron Water coolingPositive22:03

    Water absorbs approximately four times more heat than air.

    “water absorbs heat four times as much as air.”

    Listen at 22:03

  13. Will Tayloron Radiator-style data-center coolingNegative22:45

    Radiator-style cooling could increase data-center costs by approximately 5%.

    “that would then increase the costs of the data center between like around 5%.”

    Listen at 22:45

  14. Will Tayloron AI capital expenditureNeutral24:30

    Chips receive approximately half of AI capital expenditure.

    “So if $100 goes into AI CapEx... $50 will go into the chips.”

    Listen at 24:30

  15. Will Tayloron AI capital expenditureNeutral24:38

    Networking receives approximately 15% of AI capital expenditure.

    “around $15 goes into networking.”

    Listen at 24:38

  16. Will Tayloron AI capital expenditureNeutral24:47

    Power infrastructure receives approximately 20% of AI capital expenditure.

    “$20 will go into power.”

    Listen at 24:47

  17. Will Tayloron ETF Shares AI infrastructure ETFPositive26:42

    ETF Shares prefers hard-asset infrastructure over rapidly replaced chips.

    “We'd rather be in that space, the hard assets, the core components that need to be built.”

    Listen at 26:42

  18. Will Tayloron AI data centersNegative30:19

    Grid power availability is the biggest AI data-center bottleneck.

    “I would say the biggest bottleneck is power, is getting power from the grid.”

    Listen at 30:19

  19. Will Tayloron Data-center developmentNegative31:07

    Data-center development realistically takes two to four years.

    “But realistically, it's a two to four year process.”

    Listen at 31:07

  20. Will Tayloron GlobalX AI infrastructure ETFPositive31:52

    An ETF is suitable when investors favor a theme but cannot identify its winning company.

    “if you really like a theme, but you're not 100% sure on which company is going to win, then an ETF is a great way to do it.”

    Listen at 31:52

  21. Will Tayloron AI infrastructurePositive33:03

    AI infrastructure assets will be built regardless of the winning model or chip.

    “They get built regardless of whose model or whose chip is going to win.”

    Listen at 33:03

  22. Will Tayloron AI infrastructurePositive34:25

    The AI infrastructure investment theme has substantial future growth remaining.

    “we're believers that this has got a long way to run.”

    Listen at 34:25

  23. Will Tayloron AI infrastructure spendingPositive34:58

    AI infrastructure spending is unlikely to stop soon.

    “it's not a case of the taps are going to get turned off anytime soon.”

    Listen at 34:58

  24. Will Tayloron AI infrastructure investmentsNeutral35:28

    AI infrastructure should be a smaller growth allocation, not a core portfolio holding.

    “This is going to be more on that kind of smaller allocation in your portfolio because it's more of a kind of a growth, a pure growth position.”

    Listen at 35:28

  25. Will Tayloron Long-term investingPositive35:37

    Long-term investors should dollar-cost average and buy during market pullbacks.

    “if you're investing over the next 10, 15 years, dollar cost average and take opportunities when they come to you in terms of market pullbacks.”

    Listen at 35:37

Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.

Weekly email

State of the Pod, every Monday

The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.

Listen to the full episode and explore every guest, topic, and moment on PodLume.