Even Alphabet Has to Borrow. What That Tells South African Finance.

What move did this tech giant pull?

Alphabet, the parent company of Google, re-entered the bond market this week, raising almost $17 billion through its biggest-ever euro-denominated offering. The deal was structured across six tranches, with the longest note maturing in 2063. It came just months after Alphabet raised nearly $32 billion in February across dollar, sterling and Swiss franc markets, its biggest-ever US dollar bond sale, which alone pulled in $20 billion and attracted peak orders of $103 billion, well above the $15 billion initially targeted.

The number that deserves your attention is not the bond size. It is the context around it. Despite holding over $90 billion in cash, Alphabet is borrowing heavily because its AI build-out requires capital at a pace that its cash reserves alone cannot support. Last week, Alphabet raised its annual capital spending forecast by $5 billion to between $180 billion and $190 billion, and said it was planning another significant increase in 2027.

To be clear: a company sitting on $90 billion in cash, generating tens of billions in profit annually, still cannot self-fund its AI ambitions. That is the defining signal of this moment in technology.

The Price Tag Is Staggering, and It Is Not Going Down

Big Tech is now expected to spend more than $700 billion on AI infrastructure this year, a sharp increase from $410 billion in 2025. The money is going into data centres, chips and energy. Nvidia (which recently crossed a valuation of $5 trillion), their latest GPU clusters cost hundreds of millions of dollars to deploy at scale. The electricity required to run a major AI model rivals the consumption of a mid-sized city. These are not software costs that scale cheaply. They are physical infrastructure costs, and they compound.

The artificial intelligence sector has already generated approximately $300 billion in debt issuance, with market participants observing emerging signs of investor exhaustion. Google is not alone in this. Microsoft, Amazon and Meta have all materially increased their capital expenditure commitments. The race has become so capital-intensive that even the wealthiest technology companies in history are turning to global debt markets to stay competitive.

The South African Context

South Africa’s banks, insurers and fintechs are all moving on AI, and the momentum is real. South African banks are moving fast on AI, with leaders like FNB, Absa, and Standard Bank already using generative and agentic AI to cut costs, reduce fraud, and improve compliance. The financial services sector is projected to add R340 billion to GDP by 2030 through AI-driven innovations.

The results on the ground are already visible. Financial crime losses declined from R3.3 billion in 2023 to R2.7 billion in 2024, thanks in part to AI-powered fraud detection and prevention measures. Capitec reported blocking 23,000 fraudulent transactions in 2024 and saving customers over R200 million using its proprietary AI systems.

Capitec has also been applying what it calls “micro-decision lending,” where something as small as a R10 airtime advance becomes a test of trust. If clients repay promptly, the system knows it can scale up the offering. This is a deliberate move away from the blunt credit-scoring models inherited from global financial hubs, and it shows how AI can be used to open doors for the millions of South Africans who would otherwise be invisible to conventional lenders.

The broader fintech market reflects this trajectory. South Africa’s fintech sector reached $981 million in 2024 and is projected to reach $3.69 billion by 2033, at a compound annual growth rate of 15.85 per cent.

The Budget Problem Nobody Is Talking About

Here is where the global signal becomes a local concern. The same AI infrastructure that Alphabet is borrowing billions to build sits at the foundation of the models South African institutions are now deploying. FNB, ABSA and Standard Bank are accessing those capabilities at a fraction of what it costs to build them. For now, that works. But as AI becomes a core competitive differentiator rather than a feature, the cost of staying current will rise.

Current constraints in the AI industry

The shortage of skilled AI professionals within South African financial institutions. Capital is one problem, talent is another. Both are structural, and neither resolves quickly.

The SARB’s Prudential Authority and the FSCA launched a voluntary survey in late 2024 to understand the AI landscape in South African financial services. Approximately 2,100 responses were received across banking, insurance, investments, payments, pensions, fintechs and lending. Of all respondents, only 10.6% reported actually using AI. That number will grow, but it illustrates how early this market still is, even as the global race accelerates.

Where the Real Edge Lies

The South African fintechs and banks that win this transition will not be the ones that try to match global spending levels. That race is already over. The edge available here is focus, not scale.

Fraud detection is arguably the most immediate opportunity, and the evidence supports it. The combination of high mobile penetration, a large unbanked population generating rich transaction data, and an urgent regulatory environment around financial crime creates a specific problem that AI solves better than anything else. AI models can monitor activities in real time, meaning instances of fraud, mismanagement or regulatory breaches can be detected and addressed far more quickly than traditional methods allow.

Lending is the second major frontier. AI’s ability to predict user behaviour has reduced churn by 25 per cent in lending platforms, according to a 2025 Deloitte survey, and AI-driven approaches have reduced default rates by 22 per cent in micro-lending. For a market where millions of South Africans remain “thin file” clients with no formal credit history, AI-powered alternative scoring is not a product improvement. It is a market creation event.

Customer service sits third, but it is not far behind. Generative AI chatbots are already reducing the cost of routine banking interactions across the big four, and the improvements are measurable.

The Honest Take

The Alphabet bond story is not primarily about Alphabet. It is about the cost curve of AI infrastructure and what happens to every institution below the top tier when that curve steepens. South African financial institutions are not going to build their own data centres or train their own foundation models. But they will be priced for the infrastructure that makes those models run, through cloud compute costs, API fees and talent competition.

The institutions that navigate this well will be the ones that pick a specific problem, go deep on it faster than the competition, and build proprietary data advantages that make their models better than generic alternatives. That is achievable on a South African budget. Trying to be everything to everyone in AI, the way a trillion-dollar company can afford to be, is not.
The race is on. The capital requirements are clear. The question for local players is not whether to run, but where to place the bet.