Data Centers, Debates⎟Tech Infrastructure's Booms, Busts And Heated Arguments

May 15, 2025

We dive into data center investments, Stargate's $500B AI infrastructure plans, and financing uncertainty. Plus, Autodesk's controversial God-themed ad and a Texas-sized take on civil engineering.

Patric Hellermann

Shub Bhattacharya

Data Centers, Debates⎟Tech Infrastructure's Booms, Busts And Heated Arguments

tl;dr

Data center construction faces uncertainty as Stargate's $500B AI infrastructure plan hits financing roadblocks

Autodesk's controversial "God" ad campaign confuses and alienates its core enterprise customers

AEC software financials show surprising gross margins with Autodesk and Bentley operating as cashflow machines

TestFit reveals product-market-sales fit and unveils civil engineering tools to optimize dirt costs

YC and safe notes spark heated debate over real value to founders and appropriate startup financing

Planning and permitting differences between Anglosphere and European construction contribute to cost variations

Let's say you want to have a permanent record of a thing that you thought or a thing that you wanted to have written down and you didn't want to rely on electricity to save that piece of information. You can't beat paper.

🎧 Listen To This Episode

Data center construction faces uncertainty as Stargate's $500B AI infrastructure plan hits financing roadblocks

The massive Stargate project, a joint venture between OpenAI, SoftBank, Oracle, and MGX, is experiencing growing pains. This ambitious $500 billion AI infrastructure play aims to create data centers and supercomputers across the United States by 2029, but recent reports suggest trouble in paradise.

SoftBank, which pledged $100 billion to the venture, is facing challenges securing additional financing from lenders, banks, private equity firms, and other potential backers. The project appears stuck in planning stages as concerns mount over rising costs and possible data center overcapacity.

We're seeing mixed signals across the industry. TD Cohen and Alis have warned that tariffs on AI hardware could drive up costs by 5% to 15%, further complicating the financing landscape. Meanwhile, Chinese AI startups like DeepSeek are undercutting established players, casting doubt over long-term profitability. Even tech giants like Microsoft and Amazon are scaling back their data center projects amid signs of slower AI-driven demand growth.

The debate centers around whether we're overbuilding or still underserving future compute needs. Some argue that the data center boom is unsustainable – humanity tends to overreact, entering a state of euphoria before eventually reverting to a mean. The traditional pattern shows industries overshooting new markets, then experiencing a correction before finding the right steady state.

Power consumption remains a critical constraint. Utilities in many regions struggle to support the massive energy requirements of these facilities. In California, for example, PG&E costs have increased by roughly 50% over the last three years. As one podcast guest noted, "This is going to become a political issue when it starts affecting constituents – people who vote."

Despite the uncertainties, major players continue investing. Amazon, Microsoft, and Google have committed $255 billion in combined AI CapEx for 2025, up from $180 billion in 2024. However, capital allocators face challenges predicting the market's trajectory. How much is hype versus sustainable demand? What portion of the market represents timeless value?

As one construction general contractor shared, they initially entered the data center space but later pulled back, discovering that:

Their existing capabilities didn't translate to data center requirements
Owners/operators refused to let them influence design optimization
Contract penalties for delays were extraordinarily severe

Data centers currently represent a crucial economic lifeline for the construction industry. Without this sector's continued growth, some experts believe the broader construction market would have entered a depression by classical economic definitions. Some general contractors now derive up to 80% of their work volume from data centers – a dangerous lack of diversification if the market cools.

Autodesk's controversial "God" ad campaign confuses and alienates its core enterprise customers

Autodesk recently launched a marketing campaign featuring actor Tony Howe as God, drawing a "playful parallel between divine creation and Autodesk design software." The reaction across the AEC industry has been decidedly mixed, with many questioning the strategy behind the controversial approach.

The campaign, titled "Let There Be Anything," appears fundamentally misaligned with Autodesk's customer base. The construction industry tends to include more religious individuals than some other sectors, making the God comparison potentially offensive to core customers. Beyond religious sensitivities, the campaign fails to address enterprise software's fundamental value propositions – how it makes businesses better, drives revenue growth, improves ROI, or increases project efficiency.

This consumer-style brand campaign represents a departure from effective enterprise marketing principles. As one podcast participant noted, "Autodesk does not lack for a voice with the customer. They lack for a solution for the customer... They don't not buy their products because they don't know about it. They don't buy their products because they haven't solved their problem."

The puzzling aspect is how such a controversial campaign made it through multiple decision-making stages at a sophisticated company. This suggests the controversy might be intentional – but to what end? Some speculate it relates to stock market positioning, as consumer-facing brands often trade at higher valuations. With activist investor pressures mounting, perhaps Autodesk seeks to position itself as a household name rather than a specialized enterprise software provider.

Regardless of intent, the campaign represents a fundamental mismatch between marketing approach and customer base. It does little to enhance the user experience with Autodesk products or strengthen relationships with core customers. As one participant observed, "It doesn't even advertise Autodesk. It's just a random old mishmash."

AEC software financials show surprising gross margins with Autodesk and Bentley operating as cashflow machines

A recent financial benchmarking analysis comparing leading AEC software companies revealed fascinating insights into their performance metrics. Autodesk, Procore, and Bentley Systems were evaluated against peer companies like Adobe, Intuit, and ServiceNow across various financial dimensions.

The revenue growth figures show impressive performance for established players. Autodesk and Bentley achieved approximately 11% growth despite their multi-billion dollar revenue scale – a remarkable accomplishment for companies of their size. Procore leads with 26% growth, though recent guidance suggests a slowdown to about 9% for the upcoming quarter.

Gross margins tell an extraordinary story about the efficiency of these businesses:

Autodesk: An astounding 91% gross margin
Bentley: 88%
Procore: 82%

These figures place all three companies solidly in premium SaaS territory, with Autodesk standing out as a gross margin powerhouse even among elite software companies.

When examining the Rule of 40 (combining growth rate and profit margin), Autodesk and Bentley approach the coveted 40 benchmark, while Procore lags with negative operating income. This highlights the profitability differences between the established platforms and the still-scaling Procore operation.

Sales and marketing intensity (S&M spend divided by revenue) reveals interesting efficiency patterns:

Autodesk: 33 cents per revenue dollar
Procore: 45 cents
Bentley: An impressively efficient 20 cents

Bentley's sales efficiency stands out, surpassing even consumer product-led growth companies like Adobe and Intuit – a remarkable achievement for an enterprise software provider.

Free cash flow as a percentage of revenue further demonstrates the financial health of these businesses:

Autodesk: 26%
Bentley: 31%
Procore: 11%

These cash flow figures place Autodesk and Bentley in the same league as Adobe and Intuit, confirming their status as highly efficient cash-generating machines.

The revenue per employee metrics provide additional perspective on operational efficiency:

Autodesk: $400,000 per employee
Procore: $290,000
Bentley: Lagging at $190,000 despite efficiency in other areas

Current market valuations show Autodesk trading at approximately 10x trailing revenue, which is within the expected range for high-quality SaaS companies with their financial profile. Combined with their 25% free cash flow margin, this makes Autodesk an attractive investment from a financial fundamentals perspective.

TestFit reveals product-market-sales fit and unveils civil engineering tools to optimize dirt costs

TestFit, the feasibility software company led by CEO Clifton Harness, has made significant strides in expanding its product capabilities and market traction. The company has moved beyond "minimum viable product" stage with its civil engineering tools, now offering integrated solutions that address both architectural and site grading optimization.

The new civil engineering capabilities focus on reducing one of construction's major cost centers: dirt. For industrial facilities and other large developments, earthwork can represent 30% of the total cost – reaching as high as 50% in regions like Atlanta. TestFit's technology enables real-time regrading of entire sites based on parameters for finished floor elevations, helping developers reduce dirt costs from the 30% range down to around 20%.

This expansion completes TestFit's offering across multiple building types. The company started in high-density multifamily development, where dirt costs primarily related to below-grade parking structures. As they expanded into garden apartments, low-density housing, and tract home planning, dirt mitigation became increasingly important. The move into industrial facilities solidified the need for comprehensive civil engineering tools.

Beyond product development, TestFit has achieved what Harness describes as "product-market-sales fit" – a step beyond mere product-market fit. This means they've not only created software that addresses market needs but have also cracked the code on efficiently selling and delivering that value. The company now has three sales team members who have each passed the million-dollar ARR mark, demonstrating the scalability of their go-to-market approach.

TestFit's organizational structure has evolved to support this growth. They've implemented a "pod" structure that pairs go-to-market leaders with engineers, creating tight feedback loops between customers (both prospects and renewals) and product development. Harness describes this as "one of the most popular updates to our org that we've done ever."

The company now employs approximately 31 people and continues to raise capital while scaling their operations. With their dirt optimization tools now available to customers who previously lacked this capability, TestFit is focused on packaging and selling their integrated solution combining architectural design and civil engineering functions.

YC and safe notes spark heated debate over real value to founders and appropriate startup financing

A spirited debate emerged around Y Combinator's value proposition and the merits of safe notes as financing instruments. The discussion revealed strong opinions about what constitutes fair and constructive early-stage funding practices.

Several podcast participants expressed skepticism about Y Combinator's standard terms, which typically require 7% equity in exchange for admission to their accelerator program and initial funding. One guest described YC as "almost predatory to early-stage impressionable entrepreneurs," arguing that the equity stake is disproportionate to the value provided. Another concern centered around the "manufacturing process" YC has become, growing from small cohorts of 20 companies to multiple cohorts with approximately 50 companies each.

The counterargument recognized YC's potential value for first-time founders without established networks: "For someone that doesn't understand what finance is or doesn't understand what a sales org is or doesn't understand marketing... maybe that would be worth 7% to have the mentors that you need in your first company."

However, critics maintained that resourceful founders should be able to discover this information independently: "All of that information is a few clicks away... you should be able to be fluent in the terms of financing. It's not that hard." This self-sufficiency in navigating early challenges was framed as an important quality for venture-backable founders.

The conversation expanded to safe notes, the convertible security popularized by Y Combinator. These instruments were characterized as "criminal" by one participant due to their lack of investor protections and certainty. Unlike traditional convertible notes with interest rates and maturity dates, safe notes typically convert to equity at a future financing with no time constraints or interim protections.

Key criticisms of safe notes included:

No defined timeframe for conversion
No interest accrual for the capital deployed
No shareholder rights during the interim period
Limited transparency into company operations prior to conversion
The potential for founders to continually raise multiple safe notes without triggering conversions

In contrast, structured convertible notes with caps, interest, and defined conversion timelines were positioned as more balanced instruments that demonstrate founder sophistication and fairness to early investors: "For all the founders that listen to this, one of the things that will show you're very mature is if you approach... [with] some minimum certainties with a conversion, like convert in the next 24 months, a cap on the conversion, and a simple interest."

The discussion highlighted how financing practices shape the venture ecosystem. Standardized, founder-favorable instruments like safe notes have reduced friction for raising capital but potentially at the cost of appropriate investor protections and company governance. As one participant concluded, "It's kind of crazy that we needed to lower the friction to giving people that don't deserve investment money so much that we standardize this thing."

Planning and permitting differences between Anglosphere and European construction contribute to cost variations

The discussion explored differences between planning and permitting systems in the Anglosphere (US, UK, Australia) versus continental Europe, examining how these variations might impact infrastructure costs and development timelines.

The conversation centered around whether "judicial planning systems" in Anglo countries result in more expensive infrastructure compared to the "corporatist systems" found in European countries. While the podcast participants weren't entirely convinced of the premise, they identified several key differences between jurisdictions.

In the UK and US, planning permissions typically include a public consultation process where neighbors and other stakeholders can object to proposed developments. These objections can lead to planning committees (often comprised of elected officials or appointed community members) making decisions that frequently delay or alter projects. As one participant noted, "Usually this planning committee consists of 70-80 year olds who might have different incentives than the developer."

The low barrier to objections in Anglo systems was highlighted as particularly problematic. In the US, virtually anyone can file objections or even lawsuits with minimal qualification requirements. This contrasts with European systems that require objections to be well-substantiated and filed during specific windows in the process. One participant explained the European approach: "The concerns need to be very well substantiated... and then there would be a commission that looks over that and decides, 'Hey, is that worth bringing up?'"

Another key difference relates to the specificity of zoning regulations. European systems, influenced by Napoleonic code, typically have very detailed zoning requirements. While changing these can be difficult, projects that comply with existing zoning face fewer obstacles from objections. As one participant explained, "As long as you're to the zoning, the objections will not be successful."

The level of detail scrutinized during the planning process also varies significantly. In Poland, for example, planners primarily focus on ridge level, building dimensions, and entrance locations. By contrast, UK planners might object to specific window placements or detailed design elements: "This window is too big. It is overlooking your neighbor on the left-hand side."

Local variations further complicate the picture. Within the US, regulations vary dramatically by location – Houston has virtually no zoning restrictions, while California requires permits and public notices for minor home renovations. Similarly, development in high-value areas like San Francisco or London faces more intense scrutiny than projects in less dense regions.

The discussion highlighted how these regulatory frameworks impact development costs and timelines, with fragmented, objection-heavy systems potentially contributing to higher infrastructure costs in Anglosphere countries through delays, uncertainty, and defensive planning.