Supporters of Marcus Endicott’s Patreon can access weekly or monthly video consultations on this topic.
By the year 2100, South Africa will run a smaller, hotter, and more domestically oriented tourism economy than the one its operators recognise today. The most plausible picture is one of roughly twelve to eighteen million international arrivals supporting six to eight percent of GDP — a sector that has not collapsed but has rebalanced, anchored on a Western Cape rebuilt around permanent desalination and hardened water systems, a Greater Kruger that functions as actively curated managed wilderness, a Drakensberg–Maloti highland corridor that benefits from the southward migration of comfortable summer climates, and a Wild Coast and Eastern Cape arc that has absorbed much of the value the KwaZulu-Natal coastline lost to managed retreat under half a metre to a metre of additional Indian Ocean.
The climate envelope makes that future inescapable. The southern African interior is already warming at roughly twice the global mean, and under the most likely emissions trajectory — somewhere between current pledges and a slightly higher pathway — the country lands at about 2.5 to 3.0 degrees Celsius of national mean warming by century's end, with interior maxima well above that. Cape Town's brush with Day Zero between 2015 and 2018 has been formally attributed to anthropogenic warming, which made the drought several times more likely than it would have been in a pre-industrial climate. Day-Zero-equivalent events that were once-in-three-centuries in the historical record are projected to become expected within a typical three-decade planning horizon. The city has already responded by treating desalination at Strandfontein, Monwabisi and the V&A as permanent infrastructure rather than emergency stopgap; by 2100, Cape Town operates on what is effectively a post-natural water system, with dual reticulation built into new developments and the Cape Flats and Atlantis aquifers managed as engineered reservoirs.
The eastern coastline tells the opposite story. Sea-level rise at Cape Town is running at roughly twice the global mean, and the KwaZulu-Natal coast — Durban, Richards Bay, and the stretch of resort and beachfront tourism plant from the Bluff to uMhlanga — faces a planning envelope of half a metre to a metre of relative rise by 2100, with episodic storm surge superimposed. The Durban storm of April 2022, which killed 544 people and caused more than seventeen billion rand in direct infrastructure damage, has been formally attributed: the rainfall was forty to a hundred percent heavier than it would have been in a cooler pre-industrial atmosphere. Subsequent disasters across KZN and the Eastern Cape during 2025, including a flood that killed more than a hundred people in OR Tambo and Amathole districts, have confirmed the trajectory. Two generations of managed retreat from the Golden Mile and the Point Waterfront is therefore the realistic centenary plan, with insurance regulators progressively withdrawing flood and storm cover from new development in mapped retreat zones.
The Cape Floristic Region — the fynbos kingdom that defines the south-western Cape and underwrites much of the Cape Town tourism brand — is the central biological asset under the most pressure. The protected-area complex spans more than a million hectares across Table Mountain, the Cederberg, the Boland, the Swartberg and Baviaanskloof, and supports something close to nine thousand vascular plant species on a footprint that is a small fraction of one percent of the earth's land surface. Peer-reviewed projections of Proteaceae range contraction of roughly thirty-six to sixty percent by mid-century are well-established, and a commitment to range loss or local extinction for twenty to forty percent of fynbos endemics by 2100 sits at the lower end of plausible. Stellenbosch and Paarl will probably no longer be the geographic centre of South African viticulture; varietals are already migrating to cooler-end-of-window sites in Hermanus, Elgin, the Outeniqua foothills and Eastern Cape uplands, the same direction of travel that Australian, Chilean and other Mediterranean wine regions are executing.
Artificial intelligence will be a structurally double-edged force. On the conservation side it has already become operational. Drone-based detection in Kruger, machine-learning training on tens of thousands of thermal images, and deployments of AI-augmented platforms in the private reserves around Sabi Sand are the current production standard. By 2100, a realistic stack includes near-total computer-vision coverage of rhino, elephant and pangolin populations; individual-animal identification by facial, horn and ear-notch features; predictive water management of dams, aquifers and desalination plants; AI dispatch on the post-Eskom grid balancing solar, wind, batteries and green hydrogen; and crowd and safety control at Robben Island, the V&A, the Table Mountain Cableway and the Kruger gates. Language-model interpretation of Khoekhoegowab and !Xun — both critically endangered, with fewer than a thousand fluent first-language speakers between them — together with translation tooling for isiZulu, isiXhosa, Sesotho and Setswana, could let community-owned heritage sites generate localised content without paying foreign agencies.
The downside is the political problem the century turns on. Tourism today directly and indirectly employs about 1.68 million people, in a country with more than eight million unemployed, a youth unemployment rate that has peaked above sixty percent, and an expanded jobless rate above forty percent. The jobs most exposed to AI substitution — tour guides, drivers, hotel front-desk staff, food-and-beverage service, housekeeping supervision, call-centre work, travel-agency intermediation — are precisely those that currently absorb low-skill young Black workers. Multilingual AI guides at Kruger gates, the Table Mountain Cableway, Robben Island ferries or the Apartheid Museum are already technically deployable; whether the state uses BBBEE concession contracts to mandate human-in-the-loop staffing ratios is the central policy lever. Value capture is the second concern. Without sovereign cloud capacity and South African-language foundation models, the country will likely export the bulk of AI-mediated tourism economic rents to American, Chinese and Gulf providers, much as it already loses fifteen to twenty-five percent on every online booking to foreign platforms.
The regional reshuffle is already underway. The Eastern Cape is the structural winner — the Wild Coast and Pondoland, Addo, the Karoo national parks, and a less-developed coastline that the Agulhas Current peels east of are all positioned to absorb investment that KZN can no longer attract, and the Coega green-hydrogen anchor near Gqeberha is a credible mid-century electrification base for desalination and electric aviation. The Northern Cape becomes a high-margin, low-volume circuit built on the Square Kilometre Array's protected dark-sky reserve, the Richtersveld and ǂKhomani cultural landscapes, and the gradual operationalisation of the Traditional and Khoi-San Leadership Act. The Drakensberg and Maloti highlands gain share as the highveld becomes the most thermally comfortable summer destination in the country. Kruger and the Greater Limpopo Transfrontier Park continue as managed wilderness, but with a recomposed species profile, shifting bushveld–savanna boundaries, and assisted-migration interventions as standard tools.
The binding constraints are political, not biophysical. South Africa has the science, the conservation institutions, and a credible demographic envelope — the population is projected to peak around the late 2050s in the mid-seventies of millions and drift downward to between sixty-five and seventy-five million by 2100, against a 2024 baseline of roughly sixty-three million. What it does not yet have is reform of the Ingonyama Trust to give households tenure security in rural KwaZulu-Natal, full operationalisation of revenue-share rules for Khoisan tourism, sector codes that require community ownership in data and intellectual property rather than only in physical infrastructure, and a sovereign AI stack capable of resisting platform extraction. Whether the country lands a sustainable tourism sector in 2100 will turn on those choices over the next two decades — on whether AI rents are domesticated or exported, on whether managed retreat is funded as adaptation or imposed as disaster response, and on whether the worker who supervises three AI guides becomes the new normal before the worker who is one of fifty is left without a job at all.