- 2 days ago
The Government has signed three Memoranda of Understanding with U.S. companies that could bring more than US$5 billion in investment to Trinidad and Tobago.
Among the proposals are two massive technology projects...
A 300-megawatt data centre in partnership with Ernst & Young...
And another AI infrastructure project beginning at 150 megawatts, with plans to eventually expand to 500 megawatts.
Government says these initiatives have the potential to create over 5,000 jobs and help diversify the economy beyond oil and gas.
But almost immediately, concerns began to emerge.
An online petition has now attracted more than 17,000 signatures, with opponents questioning whether Trinidad and Tobago can support facilities that could require enormous amounts of electricity and water for cooling, while raising concerns about environmental impacts and the lack of public consultation.
Joining us to help unpack both the opportunities and the concerns is Executive Director of the Artificial Intelligence Innovation Centre, at the UWI Dr. Craig Ramlal.
Among the proposals are two massive technology projects...
A 300-megawatt data centre in partnership with Ernst & Young...
And another AI infrastructure project beginning at 150 megawatts, with plans to eventually expand to 500 megawatts.
Government says these initiatives have the potential to create over 5,000 jobs and help diversify the economy beyond oil and gas.
But almost immediately, concerns began to emerge.
An online petition has now attracted more than 17,000 signatures, with opponents questioning whether Trinidad and Tobago can support facilities that could require enormous amounts of electricity and water for cooling, while raising concerns about environmental impacts and the lack of public consultation.
Joining us to help unpack both the opportunities and the concerns is Executive Director of the Artificial Intelligence Innovation Centre, at the UWI Dr. Craig Ramlal.
Category
🗞
NewsTranscript
00:00More than 17,000 people have signed a petition demanding greater transparency over government's plans for AI data centers in
00:09TNT.
00:10Supporters say the projects could bring billions of dollars in investment and thousands of jobs.
00:16Critics warn they could place unprecedented pressure on a water supply, electricity grid and the environment.
00:23So is TNT on the verge of becoming an AI hub for the region or are we rushing into a
00:29project without asking the hard questions?
00:32Tonight on Beyond the Headlines, we examine the debate.
00:5390% of your immune system resides in your gastrointestinal tract.
00:57Don't let digestive discomfort hold you back.
00:59Take control of your health with Probiotic XL.
01:02Hurry, available only while stocks last.
01:06Don't miss out on feeling better now.
01:26The government has signed three memoranda of understanding with U.S. companies that could bring more than U.S. $5
01:34billion in investment to TNT.
01:36Among the proposals are two massive technology projects, a 300-megawatt data center in partnership with Ernst & Young and
01:45another AI infrastructure project beginning at 150 megawatts with plans to eventually expand to 500 megawatts.
01:54Government says these initiatives have the potential to create over 5,000 jobs and help diversify the economy beyond oil
02:03and gas.
02:03But almost immediately, concerns began to emerge.
02:07An online petition has now attracted more than 17,000 signatures, with opponents questioning whether Trinidad and Tobago can support
02:16facilities that could require enormous amounts of electricity and water for cooling, while raising concerns about environmental impacts and the
02:26lack of public consultation.
02:28Joining us to help unpack both opportunities and concerns is Executive Director of the Artificial Intelligence Innovation Center at the
02:37University of the West Indies, Dr. Craig Ramlal.
02:40Dr. Ramlal, thank you so much for joining us, and welcome to Beyond the Headlines.
02:45Thank you very much for having me.
02:46So, before we get into the debate, can you help viewers understand exactly what we are talking about?
02:54When people hear AI data center, what exactly does it mean and how does it differ from a data center?
03:04Okay, so usually a data center refers to a computation facility that will host things like websites, banking systems, databases,
03:15emails, government records, and so on.
03:19And these are quite typically your normal, ordinary cloud services.
03:23AI data centers are quite different.
03:25They are designed primarily for computation of intensive AI-based workloads, and so what happens is that the infrastructure inside
03:34of these data centers are different.
03:36The data center will have a high reliance on storage-based capacity as well as CPUs, and an AI-based
03:42infrastructure will have a high reliance on GPU or TPU types of technologies there.
03:48That's why it is so computationally expensive as well as energy mapping.
03:52Now, CPU, GPU. Is GPU generative? Could you just break that down a little bit, the jargon?
03:59Yes, thank you very much. CPU means central processing unit, and GPU means graphics processing unit.
04:06And I thank you very much for that. That actually segues into something I wanted to discuss.
04:11But actually, we should not conflate generative AI with all of AI.
04:17The artificial intelligence field has multiple things in it.
04:21You have machine learning.
04:22You have narrow-based X-ray systems that do not require large-scale types of compute.
04:28So, graphics processing units is GPUs.
04:31So, we're seeing a lot of the concerns expressed over AI data centers, even though there were two in the
04:39memoranda of understanding, a data center and an AI data center.
04:42Why would people be more concerned about an AI data center as opposed to a regular data center?
04:48An AI data center typically consumes more energy.
04:52This is because the expensive hardware that we've talked about, the GPUs, must operate at high utilization.
04:58This is how you become profitable as a data center.
05:02So, that means that they do not remain idle at all.
05:05So, on your grid, it looks like a base load of a certain amount.
05:10So, if you add this 150 megawatts, it's going to be 150 megawatts of base load.
05:15So, because of that, you have a lot of heat that is generated because you require a significant amount of
05:22energy.
05:23And then that heat, you must get rid of it.
05:26So, you have to have electricity for pumps, fans, chillers in Suriname.
05:30Right.
05:31And just to put this all into a regional context, we do have data centers within the region.
05:37I touched base with someone in Suriname, Data Sur.
05:40But do we have AI data centers within the region?
05:45We do have very small AI data centers in the region that are mainly owned by private sector companies.
05:50But nothing of this scale.
05:53This is a natural type of AI data center.
05:57So, this is something.
06:00It is equivalent to the energy required for a small city in the United States.
06:07Small city in the U.S.
06:09So, government has announced this proposal ranging from 150 megawatts to potentially 500 megawatts.
06:16And that's alongside another proposed 300 megawatt facility.
06:22Just to put that into perspective, how significant are projects of that size?
06:27This is, yes, this is quite a national scale.
06:31For our small island developing states, this is a national scale type of initiative that will most likely, so we
06:38don't know what, this is just a memorandum of understanding.
06:41And you have to understand that this is one of the lower level binding tiers of legal instruments.
06:46So, there are many things that can change right now.
06:48But this, if it were to happen, this could be something of a significant size that most likely will host
06:56international types of processing.
06:58So, maybe they will have SLAs with private companies, large scale companies like Netflix and Amazon and so on.
07:04So, this is something to assume to have.
07:08And if we could put that into perspective, I didn't actually look up this part.
07:14But if you could clarify, let's just say all three facilities or all two facilities come to fruition.
07:26Would that represent a percentage of our grid?
07:29All right.
07:31It's a very good question.
07:32What we need to recognize as well is that these systems will, if it is that they are going to
07:39be processing, especially international types of data or even local types of data,
07:46there will most likely be an SLA that requires an uptime of 99.999 something percent.
07:51So, then it's not, I understand the question of energy, if the energy systems is there.
07:57But a major question is if the energy dependability is there.
08:01If we are, from our grid, going to give them that amount of energy, most likely, and there has been
08:07a transition now with most data centers, they are looking at renewable types of energy systems.
08:11We shipped with their systems.
08:15And the reason for this is you need to keep a certain amount of uptime and service level agreements with
08:21who you are selling these data partners to, these data processing to.
08:27Right?
08:28So, it's not, it's, it's not, it's, we don't know if, if it is that they will be on the
08:33grid or not.
08:33And I don't think that that has been negotiated as yet as well.
08:36But it is most likely that they will come with some sort of energy backup or some sort of core
08:42energy that would help feed their system, that would reduce the amount of dependency on our grid.
08:49But, again, all of this is still in negotiation.
08:53But what, this amount of energy, what does it represent as a percentage of our grid?
09:03It's going to be a high percentage of our grid, if we were to take it directly from the grid.
09:09So, but you remember, again, we are an industrial-based economy.
09:13And because of that, we have built infrastructure that handles this type of processing.
09:19But, again, I want to make it clear that it is, it is most likely that they may ship with
09:24their own energy systems.
09:25In order to be a tearful facility, you must have redundancy in your energy systems and your cooling systems.
09:32So, I mean, it's going to be, they have to, yeah, no, no, no, nothing like that, no, no.
09:38This is like international partners, too, that you are signing service level agreements with, so, yeah.
09:43Is that a positive for our country to have that type of technology here?
09:50Okay, but now we're getting to the meat of the questions.
09:54And what I would say is, it depends on how we negotiate it.
09:58So, if we can negotiate a way that we know that it will produce jobs, so, we've heard the numbers,
10:055,000 jobs and so on.
10:06And what we want to do is we want to expand our negotiation.
10:10Beyond that, we want to give access to SMEs, to startups, to research centers, to universities that will end up
10:18building an AI ecosystem from this type of capacity online.
10:21So, if you say something like, for every one megawatt, you will give access to these other partners, 0.2%,
10:280.2% of that one megawatt, then something like that works out.
10:31Because what we do then is we get involved in the knowledge economy, which is what the world is moving
10:37towards.
10:37And so, then we can develop products and services at scale that we can then export to other markets outside
10:44and bring better FDIs into our country.
10:47Now, you did touch briefly on jobs.
10:49And I'm hearing my producer, Candy, asking, does our workforce have the requisite skill set?
10:58And this may be a good point for you to put in a plug for the Artificial Intelligence Innovation Center,
11:03Dewey.
11:04Yes, let's discuss it.
11:06So, on Monday, I would have been on the morning show, on TV6 as well, discussing the University of the
11:12West Indies' new Masters of Applied Science in AI, our MPhil in AI, and PhD in AI programs.
11:20So, I mean, we, okay, this is all in coincidence.
11:23We did not know that all of these things were happening at the same time.
11:26We were building these programs out since 2023.
11:30So, this has been in the pipeline for a very long time.
11:34And I think that if you are looking at bettering your AI skills, you are looking at, you know, operating
11:41some of these equipment,
11:42or even getting to an advanced level where you can start up your own company and look at how you
11:48can get involved in export markets and other markets for AI,
11:51I think you should look at registering for these programs.
11:55The deadline for the programs is the end of July.
11:57July 25th, I think.
11:59So, please try to, yeah.
12:00So, but back to the question, will, let's say, I don't know how long it'll take for these projects to
12:09actually come into fruition,
12:10but would our workforce have that type of training?
12:16What type of jobs would these 5,000 jobs be?
12:19All right.
12:20Yeah, thanks.
12:21Okay, let's separate it into two timescales.
12:24The first timescale is where we currently are.
12:26So, the Artificial Intelligence Innovation Center, and I will now insert my plug, is the largest AI center in the
12:33Caribbean.
12:34We have about 73 members that exist from multiple universities and multiple parts in the Caribbean,
12:40places like Cambridge University, Oxford, MIT as well.
12:42And we work with private sector.
12:45We write policies for the Caribbean, and we also do these academic programs that I talked about so far.
12:51Before all of these news, what we were doing would have been partnering with private sector.
12:56So, that we would build products and services that private sector themselves can be economically benefit the private sector to
13:05do, right?
13:06So, we'll go with a company.
13:07We will talk about the problems that they have.
13:09And then we will start building out these AI systems.
13:12From that, our programs will then feed our students into those companies so that you have a talent pipeline that
13:20runs all the way from us starting off with building the products and services.
13:24So, individuals who know how to build on top of those systems or to even operate those systems, and you
13:31have a flourishing type of structure there that forms this kind of nice supply chain because you will come back
13:37to the AI center to look at more and deeper research that you want to do with AI.
13:42Now that we have, or may have, this particular AI data center discussion on board, you are now operating at
13:53scale.
13:54300 to 500 megawatts is no, like I said before, this is no joke.
13:58This is massive international scale type of compute facilities that will be available.
14:04If it is available to the SMEs, to the private sector, to the researchers, to public sector, then we can
14:10do multiple things on top of it because AI is in all of our sectors, right?
14:15So, they would help as a huge productivity boost if we do it particularly right.
14:20So, I think that that would be a chalk and cheese kind of scenario.
14:23Chalk scenario, which is before, where we do these things and have it organically grow, and the private sector grows
14:30with it.
14:30And then cheese is where, you know, we have access to massive compute exactly, and what can we do with
14:36this kind of compute?
14:38I just want to go back a little bit.
14:41We did talk about water.
14:43You did explain the need for water.
14:45These power plants, these data centers generate a lot of heat, so you need water to help cool down.
14:50Now, the prime minister did explain that non-potable water will be used.
14:56Does that clear up any concerns that persons would have raised?
15:01I think so.
15:03I think, first of all, we need to step back as a society, and we don't have anything pen to
15:11paper to know what is the water requirements that is required for these types of data centers.
15:16First of all, we are in a high-humid environment in the tropics.
15:20We don't know if evaporated cooling will work as efficiently to warrant water cooling for us.
15:25We don't know if air cooling alone should be used.
15:27We don't know if there is what they call a closed system, a closed-loop system, where you would pass
15:34the water net to the GPUs and then have air that cools the water.
15:39That you always need some kind of thin or fluid, right?
15:42So it's either air or some liquid-like water.
15:45But I think that that's a step back.
15:48We don't know.
15:49And when we do know, then I think we can make a decision.
15:53But there have been operations and discussions that companies are not using potable types of water.
16:01It is possible to use effluent type of water instead if it is that it meets the ASHRAE levels of
16:07certification.
16:08So it is quite possible that with minimal treating, we can use non-potable water.
16:12But we have to see.
16:14We have to still see if we are going to use water at all in cooling these particular systems.
16:18You did mention there are alternative cooling technologies.
16:22Yes.
16:22Yes.
16:22And these plants come with that.
16:24They come with their plans often until, as you mentioned, alternative sources of energy, not necessarily from our grid.
16:34But beyond water and electricity, we keep on hearing about environmental impacts.
16:40What environmental impacts would persons be thinking about, should they be thinking about?
16:46Yes.
16:46So there are a few.
16:48So there will be, everyone immediately runs towards water and they run towards greenhouse emissions, right?
16:56But, of course, we just discussed that.
16:58That is, if you are going to be using gas turbines or diesel generators or if you are going to
17:04use water from potable sources and so on.
17:06So we discussed that.
17:07Let's remove that to one side.
17:09The other two are going to be noise pollution.
17:12And these are generated because of the fans that they use for cooling.
17:17That's one.
17:18And the other one is going to be, how are you going to dispose of the silicon and the GPUs
17:23and these special materials that you need for our racks and our processing?
17:30So I think that both of them will have to have special regulations in how you deal with that.
17:36The location is going to be very important for the noise issue.
17:39But we are a country, we are an industrialized country, so we do have industrial parts.
17:44We know how to deal with this already.
17:46We have been doing this for decades.
17:49The disposal of the e-waste is one that we might have to ramp up.
17:53I know that the EMA has some regulations of doing that, but this may require some more special type of
17:58regulation that we will end up dealing.
18:00Could this also be an opportunity for organizations like the scrap iron dealers to get into e-waste?
18:09I mean, or is it highly technical?
18:15I don't know.
18:19It could be.
18:20I'm not sure.
18:21It depends on what the EMA sets aside.
18:23But would locating these facilities in existing industrial estates, you know, like Point Lisas, reduce some of those environmental risks?
18:32Yeah, these are zones, right?
18:34So you would be reduced to especially noise.
18:38We must remember, so you raise Point Lisas, they have their own water facility.
18:43So they have desalpha that provides water into the industrial zone there as well.
18:48So things like this can definitely help with some of the alleviating some of the issues.
18:52If we are really concerned with carbon, if we are really concerned with greenhouse gases, we must recognize that around
18:59the world, people have been powering 300 megawatt type size data centers on renewables, like hydrogen, like solar, like wind,
19:09or some mix of those renewable sources.
19:12So those things do exist.
19:14And it depends on how we negotiate and the private sector on what they build and what land they require
19:21to do these particular things.
19:23Now, I was chatting, Dr. Amlal, with an executive from Data Sur.
19:28That's, of course, a Surinamese data firm.
19:30So what the executive was telling me is that there are companies, as you mentioned as well, that already have
19:37in-house data centers.
19:38So this is not a foreign concept at all.
19:44What she was saying is that a local data firm could reduce some of those electricity and water concerns.
19:52How do you feel about that?
19:55And then we don't know if these firms will be for local data or them getting the impression it is
20:03for international data.
20:05Yes, that is a very good question.
20:07I think that that is a discussion on consolidation of compute as well as co-location for a cloud facility.
20:16The nail in the coffin there will definitely be the use cases of these types of systems and the service
20:23level agreements for uptime.
20:25So what normally happens is that for smaller private sector-based initiatives, you will have spaces of idle time that
20:33you would be running your data centers for.
20:35So there will be cases where you turn it off.
20:36You don't need to run it, so you turn it off.
20:38So what they do is they come together and they try to move around compute so that they essentially have
20:46some sort of uptime within all of the data center systems.
20:50So you have less servers, and so your energy requirements is less, and your water requirements for cooling is less
20:56because you have less data centers that have no one uptime because your workloads are not 24-7, 365, and
21:03so on.
21:04So this is not the case. This is a massive data center built for an international market that's most likely
21:11all of the world's data might be coming into here.
21:15We are not sure. We don't know for sure, but of this size, it is typical that that normally happens.
21:20What sectors locally could tap in if there is a local portion that will be assigned to private sector or
21:29even state?
21:30What are some of these sectors that would need or have use for data farms?
21:38I mean, just come into mind, I just thought about governments and banks. I'm not sure anyone else.
21:44Yeah, I would say everyone. So I would say, so there would be agriculture for processing on a certain amount
21:51of competition for meteorology and climate and that kind of thing.
21:56You would have things like cybersecurity professionals would use this kind of information, industry for oil and gas, research-based
22:03companies.
22:06You know, it goes, the list will go on and on. We are, we have to remember, right, what AI
22:12is. AI is a knowledge manipulation tool, especially generative AI systems.
22:17And then you have other types of AI tools being there. But this generative AI system, if you are looking
22:25to process that kind of workload, all sectors will be using it.
22:31That's why we have these conversations all the time with all of these sectors being involved there.
22:35So instead of looking at the sectors like banking and so on, you have to look at it like finance.
22:39You have to look at it like, you know, public administration, because that is what comes to mind right now,
22:44oil and gas and so on, energy and so on going on.
22:47So, everyone will, if we do it right, if we negotiate right, everyone should be able to benefit from these
22:55types of services.
22:58Internationally, who usually uses these AI data farms?
23:06Okay, so you would have your hyperscales that would have service level agreements with the companies who own the infrastructure
23:15running it.
23:15So you would have the Amazons, the Netflixes, the Googles, the Microsofts, the OpenAIs, all of those guys you would
23:25most likely have running these things.
23:27The types of data that you would normally get would be for those AI workloads.
23:32So you can look at the search for maybe CharGPT, for Claude, generative AI types tools, and so on, we'll
23:40be using these types of information.
23:42Well, these types of data centers, cybersecurity, data residency, all of these things would be at least.
23:48Our data protection will have to be quite strong for us to get into this.
23:52So, one of these farms essentially would be a mix of all those probably sectors or companies tapping into this
24:01particular data farm,
24:02or would it be solely owned by Google or Amazon?
24:06We are not sure because we, again, this is still a negotiation phase, but what they normally do is they
24:14own portions of the service level.
24:19So how the workloads are appropriated and moved is something that jumps around from geography to geography, right?
24:27It depends.
24:28Remember, we are in the Caribbean.
24:30This is a very good place to be for a data center because we have connections, close proximities geographically to
24:37North America and South America.
24:39So we can process workloads from North America and South America very quickly because the subsea cables will allow us
24:44to have a very low latency in data transmission between these two spaces.
24:49So you might get a Google, you might get a Netflix, and then you would be processing information for the
24:55South American markets,
24:56the Caribbean markets, Central American markets, and North American markets.
24:59So I was about to ask, and this was one of my questions, why would the United States, with their
25:06own large land space and favorable energy costs,
25:10why would they want to come to Trinidad and Tobago of all the countries in the world to set up
25:15this data farm and the AI data farm?
25:19Because they probably like us.
25:20No, I'm kidding.
25:20That's not the answer.
25:22It's probably the politically correct answer.
25:25So let me tell you, I think we are very strategic, I would say, in one, where we are located
25:35because of the geolocation that I just mentioned.
25:39We do have an established natural gas and power generation infrastructure, if we were to go to that way.
25:44We do know how to produce industrial land, so we know how to develop that land and how to have
25:53our environmental zoning around that land or regulations around that land.
25:57We can create special economic zones.
26:00And then, of course, there's the political and the commercial relationships that we have with the United States.
26:06So I think you should not look at it like, they have land space, yet they are not doing it
26:13there.
26:14Why aren't they doing it there?
26:17It's multiple things.
26:18We have a talent pool.
26:19We have a very high human capital that we can put into these data centers to run these particular systems.
26:27It's very, very strategic.
26:28If I were there, I would have, yeah, looked at it like from that angle as well.
26:33If these projects are to proceed, what safeguards would you want to see in place?
26:39Okay, so we need to really have discussions around the environment safety.
26:44That's a big one.
26:45We need to know where is the energy coming from, right?
26:50Is it that the gas is coming from our grid or where is that energy coming from?
26:54Where is, how are you pulling it?
26:56Where is the water coming from?
26:57Where is it located so that we know how the noise is going to be dealt with?
27:01I think another question on data sovereignty and data processing is a big one.
27:07We should know if we are taking data from international markets,
27:12if it is that our own data rights and data protection allows us to process that particular data in a
27:18first place,
27:19and if their data rights allow them to move that data out of their jurisdiction into our jurisdiction.
27:24I think that's a big two that I would be thinking about before we get into any kind of agreement.
27:31And then, very importantly, how does our public benefit beyond the creation of jobs?
27:37We need to benefit in such a way that this grows beyond...
27:41So, if we lose the data center tomorrow, we have it today, we lose it tomorrow.
27:44We need to be in a position that de-risks us from not having that data center at all.
27:51So, did we create the startups?
27:52We'll be able to mobilize certain services and products to export markets so that we bring in that FBI,
27:58and then we can use data centers outside.
28:01So, I think that's the main three.
28:02It kind of feels like a little Silicon Valley here, a little bit, you know.
28:07But my final question, 30 seconds, if you were advising Cabinet today,
28:11what are the three questions you believe government must absolutely answer before these projects materialize or advance?
28:19I would pose it as what those questions the government must ask the private sector so that they can answer
28:24all of the people to have.
28:26And those are the three I called before.
28:28Where will every megawatt come from and a litre of water come from?
28:31What measurable value will remain in Trinidad and Tobago beyond the existence of a data center and talent that we
28:37will have in the data center?
28:38And then, you know, what protects us if some of these forecasts go wrong, right?
28:44What is the backup plan?
28:45Thank you very much, Dr. Craig Ramlau.
28:47Thank you for joining us.
28:48The debate over AI infrastructure is unlikely to end anytime soon.
28:53For supporters, these projects represent a once-in-a-generation opportunity to diversify the country's economy
29:00and create thousands of high-value jobs.
29:03For critics, they raise important questions about sustainability, transparency,
29:07and whether the country's water and electricity infrastructure can support developments on that scale.
29:13As these proposals move through the due diligence phase,
29:17many will be watching to see whether those questions are answered before any final decision is made.
29:23I'm Urvashi Tawari Rupnarai and thank you for watching Beyond the Headlines.
Comments