AI's debt colony

In 1528, the Welser banking family of Augsburg struck a deal with Emperor Charles V: in exchange for debts the Emperor couldn't repay, they'd receive the Province of Venezuela to colonise and exploit. They called it Klein-Venedig — Little Venice. It lasted eighteen years, and it's one of the most instructive failures in colonial history.

The credit trap

The Welser ran Venezuela as a credit colony. Settlers arrived with nothing and were forced to buy provisions — food, horses, arms — on credit from the Welser themselves. Within months, the entire population was indebted to its overlords.

This changed everything. The debt created a desperate, extractive logic that consumed the colony from within. Sustainable industries like balsam processing were ignored — too slow, too capital-intensive — in favour of speculative entradas searching for gold that didn't exist. The encomienda system, which sustained other Spanish colonies, was never established. No farms were built. No commerce developed. Everyone was so consumed by the need to find quick returns that a functioning settlement never materialised.

By 1546, the colony had nothing to show for itself except enslaved natives and mounting debts. The Spanish governor executed the Welser representatives. Charles revoked the charter. The venture that was supposed to generate wealth had instead consumed every resource — human and natural — in a frantic scramble to service the debt that preceded any actual economic foundation.

The debt didn't just fund the colony. It shaped it. It selected for short-termism, recklessness, and exploitation — not because the colonists were uniquely bad people, but because that's what the financial structure demanded.

The new entradas

The parallels to the current AI buildout are hard to ignore.

The hyperscalers are projected to spend $650–690 billion on capital expenditure in 2026, nearly doubling 2025. Roughly 75% of that — around $450 billion — is going directly to AI infrastructure. Amazon alone plans $200 billion in data centre spending this year, which Morgan Stanley estimates will push it to negative free cash flow of $17–28 billion. Alphabet's free cash flow is projected to drop 90%. Meta's the same.

To bridge the gap, they're borrowing. Goldman Sachs found that hyperscalers have taken on $121 billion in new debt in the past year — a 300% increase from typical levels. JP Morgan estimates up to $7 trillion of AI spending will ultimately be financed with borrowed money. Morgan Stanley and JP Morgan project the tech sector may need to issue $1.5 trillion in new debt over the next few years.

This is not investment backed by proven returns. A 2025 MIT report found that 95% of organisations investing in generative AI are getting zero return. Bain estimates the industry needs $2 trillion in annual AI revenue by 2030 just to justify the capital already committed — more than the combined 2024 revenue of Amazon, Apple, Alphabet, Microsoft, Meta, and Nvidia.

The money is flowing before the gold has been found.

What debt demands

Just as with the Welser colony, the debt isn't neutral. It's shaping what gets built and what gets sacrificed.

What gets sacrificed is caution. OpenAI has disbanded two successive safety teams — its Superalignment team in 2024 and its Mission Alignment team in February 2026. Jan Leike, who left for Anthropic, wrote publicly that "safety culture and processes have taken a backseat to shiny products." Miles Brundage, who led OpenAI's AGI Readiness team, concluded on his departure: "Neither OpenAI nor any other frontier lab is ready." Anthropic's own head of Safeguards Research resigned, writing that he had "repeatedly seen how hard it is to truly let our values govern our actions."

OpenAI quietly removed the word "safely" from its mission statement. It fired a safety executive who opposed its pornographic content rollout. Within a single recent week, senior safety researchers at OpenAI, Anthropic, and xAI all resigned with public warnings.

This isn't a coincidence. This is what debt demands. When you've committed hundreds of billions before the technology has proven itself, you cannot afford to slow down. Safety research that might delay deployment becomes an existential threat — not to humanity, but to the balance sheet. The financial structure selects against caution, just as it did in Klein-Venedig.

What kind of colony are we building?

The Welser colonists weren't searching for gold because it was a good strategy. They were searching for gold because the debt gave them no other option. The financial structure made thoughtful, sustainable development impossible.

The question for AI isn't whether the technology will be transformative — it may well be. The question is what version of it gets built when the builders are $7 trillion in debt before the thing works. When the financial pressure demands growth at all costs, what gets shipped? What corners get cut? What warnings get ignored?

The Welser colony didn't fail because Venezuela lacked resources. It failed because the debt ensured those resources could never be developed properly. The colony needed patient capital and got extractive credit. It needed settlers and got treasure hunters.

We might be making the same trade.