OpenAI’s choice of Israel for its first commercial outpost outside the United States clearly shows where money and attention are going. It is not going where governments boast on stage. It is going where startups already buy, engineers use the tools, and private capital has already done half the job before the sales team arrives.
This should be a stark realization for South Africa. While global AI vendors target a market with strong adoption, dense talent, and billions already flowing through it, we still debate electricity, broadband, and whether the next round of public discussions will benefit a township school or a small business in Soweto.
Israel got there first
OpenAI’s move into Israel is commercial first, technical later. There is no development center yet, no grand engineering campus. It’s patient business development that precedes bigger bets, with local hires, enterprise sales, and relationship-building within the country’s high-tech sector.
A senior Israeli executive, formerly running AWS for Startups in Israel, has already joined OpenAI with a regional remit that still focuses on Israel. OpenAI is also recruiting an account director for the Israeli tech market, formally based in Paris but expected to work from Israel and report into Europe. If the operation grows, a country manager will likely follow.
This is the pattern: sales first, engineering second. India followed the same route. Google’s Israeli operation began with business activity before deepening into R&D. Nvidia acquired stronger local capability through Mellanox. Amazon, Microsoft, Apple, Google, and Nvidia all have significant engineering footprints there already. OpenAI is late to the party, but it has arrived where the table is already set.
Israel has also become one of the fastest adopters of generative AI. Anthropic reports it ranks first in per-capita Claude usage among working people. A Bank of America study, reported by Calcalist, placed the country outside the very top tier overall because national infrastructure and public investment lag, but the private sector numbers are compelling. LinkedIn data in that same report put Israel first in AI talent concentration and third in AI investment between 2013 and 2025, attracting about $19 billion.
This is the real story. AI companies chase markets where customers understand the product and the talent pool knows how to build around it.
South Africa keeps missing that train
Here, the conversation is still too often inflated beyond what the ground supports. We have pockets of real technical talent, some sharp startups, and a few companies doing serious work. However, a serious AI hub is not built on a handful of clever founders and a panel discussion in Sandton. It rests on reliable power, affordable networks, schools that produce maths and engineering graduates at scale, and a state that invests instead of performing concern.
We lack that base. Load shedding may come and go in the headlines, but its damage affects every office backup plan, every small business invoice, and every student trying to load a course on unstable data. Internet access remains patchy and expensive in places where it should be ordinary. The gap between a suburb with fiber and a township with one unreliable signal bar is not a technical detail. It is the difference between being able to work with AI tools and being shut out of them.
The result is predictable. The people most able to benefit from AI are already the people most likely to have devices, connectivity, private schooling, and English fluency polished for global platforms. The rest are told to be patient while the future arrives.
The divide will not stay digital
AI will sort people, not just automate abstract office work. Firms that can buy better tools will work faster, market better, and hire smarter. Workers without access will be expected to absorb the shocks. This hits call centers, admin jobs, routine clerical work, and some manufacturing functions first. These are not elite jobs; they are the jobs that keep households afloat.
The class line is obvious. The racial line is just as obvious, because the old geography of inequality still runs through access to schools, transport, connectivity, and disposable income. If AI becomes another layer of advantage, it will not distribute itself evenly into Khayelitsha, Hammanskraal, Mdantsane, or rural Limpopo just because the brochures use the word inclusion.
There is also a quieter loss. Imported AI systems are built for the markets that funded them. They do not automatically understand South African accents, township business patterns, informal trade, local health realities, or the practical mess of public services. If we only consume tools built elsewhere, we end up paying for someone else’s priorities and then pretending that is progress.
The real race is not being run here
South Africa likes to describe itself as a player in the global AI story. That is generous bordering on fantasy. A player needs scale, capital, research depth, and institutions that can back the sector for years. What we have instead is a fragile base and a large appetite for public relations language.
If the country wants a real place in this economy, the first job is boring and expensive. Build broadband that reaches beyond the affluent edge of the map. Keep the lights on. Fix the public school pipeline so maths, coding, and data skills are not reserved for private institutions. Fund serious R&D instead of sprinkling grants like confetti. Back local startups with capital that does more than host a launch event.
Without that, the gap widens. The rich will use AI to move faster. The poor will meet it as another filter they did not design and cannot afford.
