Dr Joel Teelucks­ingh

There is a strange irony in the na­tion­al con­ver­sa­tion now un­fold­ing. We are be­ing asked to con­sid­er host­ing ar­ti­fi­cial in­tel­li­gence (AI) da­ta cen­tres but most of us know re­mark­ably lit­tle about what they are, what they con­sume or what they may de­mand of the coun­try. As a re­sult, most of us turned to ar­ti­fi­cial in­tel­li­gence to in­ves­ti­gate the in­fra­struc­ture that makes the tech­nol­o­gy pos­si­ble.

“The cloud” sounds weight­less and spir­i­tu­al. Our pic­tures, emails, X-rays, school as­sign­ments, What­sApp back­ups, hos­pi­tal sys­tems, gov­ern­ment files and late-night search­es all drift some­where above us. This cloud is not in the sky.

It sits in build­ings filled with servers, ca­bles, se­cu­ri­ty sys­tems, back­up gen­er­a­tors, cool­ing tow­ers, elec­tri­cal sub­sta­tions and enough hum­ming ma­chin­ery to make the fu­ture sound like a bee­hive. It has a pow­er bill and needs wa­ter, land, roads, per­mits, en­gi­neers, tech­ni­cians, guards, con­tracts, fi­bre-op­tic ca­bles, fire sup­pres­sion sys­tems and po­lit­i­cal de­ci­sions.

We should re­spond with a na­tion­al ques­tion:

What ex­act­ly are we be­ing asked to host?

The promise is ob­vi­ous. We have spent decades speak­ing about di­ver­si­fi­ca­tion. Oil and gas have fund­ed roads, schools, hos­pi­tals, sub­si­dies, salaries and na­tion­al com­fort, but they have al­so made us vul­ner­a­ble to price shocks, pro­duc­tion de­cline and po­lit­i­cal pro­cras­ti­na­tion.

AI da­ta cen­tres ar­rive with the lure of tech­nol­o­gy, for­eign in­vest­ment, jobs, in­fra­struc­ture, glob­al rel­e­vance, dig­i­tal trans­for­ma­tion and a chance to move be­yond ex­port­ing raw en­er­gy to­ward pow­er­ing in­tel­li­gence it­self.

There are gen­uine ad­van­tages worth con­sid­er­ing.

A well-de­signed da­ta-cen­tre in­dus­try could bring in­vest­ment, tech­ni­cal train­ing, con­struc­tion work, en­gi­neer­ing jobs, cy­ber­se­cu­ri­ty ca­pac­i­ty, fi­bre-op­tic up­grades and pres­sure to mod­ernise the na­tion­al grid. It could help po­si­tion the coun­try as a re­gion­al dig­i­tal hub. It could sup­port cloud ser­vices for Caribbean banks, uni­ver­si­ties, hos­pi­tals, cre­ative in­dus­tries and gov­ern­ments. It could en­cour­age young peo­ple to see fu­tures in soft­ware, elec­tri­cal en­gi­neer­ing, cool­ing sys­tems, da­ta sci­ence, ro­bot­ics, cy­ber­se­cu­ri­ty and ar­ti­fi­cial in­tel­li­gence with­out feel­ing that mi­gra­tion is the on­ly path to rel­e­vance.

In health­care alone, the pos­si­bil­i­ties are re­al. AI can help read scans, or­gan­ise records, de­tect pat­terns in lab­o­ra­to­ry da­ta, sup­port telemed­i­cine, im­prove pub­lic health sur­veil­lance and as­sist re­search. A Caribbean dig­i­tal in­fra­struc­ture, prop­er­ly gov­erned, could sup­port lo­cal med­ical da­ta, lo­cal dis­ease pat­terns and lo­cal in­no­va­tion in­stead of leav­ing every­thing to dis­tant servers and im­port­ed as­sump­tions.

But here is the prob­lem.

A large AI da­ta cen­tre is an in­dus­tri­al fa­cil­i­ty. It con­sumes elec­tric­i­ty at a scale most cit­i­zens have nev­er had to imag­ine. It may re­quire large vol­umes of wa­ter de­pend­ing on its cool­ing sys­tem. It pro­duces heat. It needs re­dun­dan­cy. It de­mands re­li­a­bil­i­ty.

We are a small is­land state with traf­fic jams af­ter one ac­ci­dent, com­mu­ni­ties that still re­ceive sched­uled wa­ter, hos­pi­tals that strug­gle with age­ing in­fra­struc­ture, flood­ing af­ter heavy rain, and cit­i­zens who know what it means to store wa­ter in tanks be­cause the tap can­not al­ways be trust­ed.

That does not mean we must re­ject da­ta cen­tres. It means we must ne­go­ti­ate like adults.

The first is­sue is elec­tric­i­ty. If AI da­ta cen­tres re­quire vast pow­er, where will that pow­er come from? Will they draw from the na­tion­al grid?

Will they build ded­i­cat­ed gen­er­a­tion? Will nat­ur­al gas be used? If so, at what price, and who ben­e­fits?

Will re­new­able en­er­gy be part of the de­sign or mere­ly part of the press re­lease?

Will or­di­nary cit­i­zens pay high­er elec­tric­i­ty costs be­cause grid up­grades are need­ed to serve pri­vate fa­cil­i­ties?

Will house­holds sub­sidise the fu­ture while still wait­ing on bet­ter roads, safer hos­pi­tals and re­li­able wa­ter?

The sec­ond is­sue is wa­ter. In a coun­try where peo­ple have be­come ac­cus­tomed to wa­ter sched­ules, any wa­ter-in­ten­sive project must face a high bar of ex­pla­na­tion. What cool­ing tech­nol­o­gy will be used? Will it use potable wa­ter, re­cy­cled wa­ter, sea­wa­ter, closed-loop sys­tems or air cool­ing? What hap­pens dur­ing drought? What hap­pens when near­by com­mu­ni­ties are al­ready frus­trat­ed? Will there be in­de­pen­dent mon­i­tor­ing of wa­ter with­drawals and dis­charge? Will da­ta be pub­lic?

The third is­sue is jobs. When­ev­er large projects are an­nounced, job num­bers ar­rive like con­fet­ti. Thou­sands of jobs are promised. But da­ta cen­tres are un­usu­al. They cre­ate many jobs dur­ing con­struc­tion, but once op­er­a­tional they are of­ten far less labour-in­ten­sive than the pub­lic imag­ines. The best per­ma­nent jobs may re­quire high­ly spe­cialised train­ing. If we are not care­ful, we may pro­vide land, pow­er, wa­ter and tax con­ces­sions while im­port­ing many of the best-paid tech­ni­cal roles.

The re­al ques­tions are: what kind of jobs, for how long, at what salary, with what train­ing path­way, and for whom? If da­ta cen­tres come, they must come with bind­ing com­mit­ments to schol­ar­ships, ap­pren­tice­ships, lo­cal pro­cure­ment, uni­ver­si­ty part­ner­ships, tech­ni­cal train­ing, in­tern­ships and trans­par­ent re­port­ing on lo­cal em­ploy­ment.

A coun­try can host servers and still have chil­dren leav­ing school dig­i­tal­ly il­lit­er­ate. It can pow­er AI mod­els and still have pa­tients car­ry­ing pa­per notes be­tween clin­ics. It can an­nounce bil­lion-dol­lar dig­i­tal projects and still have cit­i­zens queue­ing for ba­sic doc­u­ments. The pres­ence of a da­ta cen­tre does not au­to­mat­i­cal­ly make a coun­try smart.

Large in­fra­struc­ture projects do not land equal­ly on all cit­i­zens. Debe, Point Lisas, La Brea, Waller­field, Tamana, Cove or wher­ev­er such projects are lo­cat­ed: com­mu­ni­ties de­serve re­spect be­fore de­ci­sions be­come ir­re­versible. Pub­lic con­sul­ta­tion should be se­ri­ous, ear­ly and in­formed.

The world is chang­ing whether we ap­prove or not. AI will af­fect med­i­cine, bank­ing, ed­u­ca­tion, law, agri­cul­ture, me­dia, se­cu­ri­ty, trans­port and em­ploy­ment. A small coun­try that re­fus­es to un­der­stand it will be shaped by it any­way, on­ly from a po­si­tion of weak­ness.

The best po­si­tion is not “yes” or “no.”

The best po­si­tion is: prove it.

Prove that the pow­er plan is sound, cit­i­zens will not sub­sidise pri­vate prof­it, wa­ter se­cu­ri­ty is pro­tect­ed, en­vi­ron­men­tal stan­dards are en­force­able, jobs will be re­al, uni­ver­si­ties will be in­clud­ed and con­tracts are trans­par­ent.

The cloud is com­ing ashore.

Be­fore we wel­come it, we should re­mem­ber what every child even­tu­al­ly learns: clouds may bring rain, but they may al­so block the light.