Sustainable Data Infrastructure: Clara Tsao on Filecoin's Vision for AI and Climate-Friendly Storage (Part 2)

Anthony Perl: 00:00:02

Sustainable data infrastructure.

2

: 00:00:04

Clara Tsao on file coin's, vision for AI

and climate friendly Storage part two.

3

: 00:00:10

In this episode, Dr.

4

: 00:00:11

Jemma Green continues her conversation

with Clara Tsao exploring how Filecoin

5

: 00:00:15

is pioneering sustainable approaches

to data storage in the age of ai.

6

: 00:00:20

From solar powered data centers to

partnerships with GPU and compute

7

: 00:00:24

networks, Clara reveals how decentralized

infrastructure can address the massive

8

: 00:00:29

energy demands of modern technology

that discuss file coin's role in

9

: 00:00:34

the AI ecosystem, the importance of

data verification, and how Clara's

10

: 00:00:38

background in national security and

disinformation shaped her vision for

11

: 00:00:42

a more transparent digital future.

12

: 00:00:44

I'm your co-host Anthony Pearl,

and whether you're an investor or

13

: 00:00:47

a startup looking for insights,

it's time to get unblocked.

14

: 00:00:54

Jemma Green: You mentioned

data centers powered by solar.

15

: 00:00:57

I mean, it's not actually physically

possible to have a data center entirely

16

: 00:01:01

powered by solar because obviously.

17

: 00:01:03

It's only generating electricity when the

sun's shining, but you could create 24

18

: 00:01:07

7 renewable power through a combination

of say, solar, wind, battery storage

19

: 00:01:12

or even renewable energy certificates

to cover the period when the wind's

20

: 00:01:17

not blowing or the sun's not shining.

21

: 00:01:19

But my question about that is, is, is

that something that's actually sought

22

: 00:01:23

after where people are looking for storage

that is a hundred percent renewable?

23

: 00:01:29

Clara Tsao: Yeah, so we actually have

a program called FCO Green where we

24

: 00:01:32

actually give, again, like I said, we

are a network and we don't really tell

25

: 00:01:36

people what to do, but if people have

really great ideas that they want to

26

: 00:01:40

have as a blueprint for others to follow

by, we definitely help incentivize.

27

: 00:01:45

Through the form of grants

to get people to get started.

28

: 00:01:48

So the example I was talking about earlier

for a solar powered data center on our

29

: 00:01:53

network is by a company called Scent.

30

: 00:01:56

They're based in Amsterdam, and I'm

sure they still use electricity like

31

: 00:02:00

you said, to some degree because

it's not 24 7 sun over there, but

32

: 00:02:05

they, they have majority of it.

33

: 00:02:07

Majority of their energy generation

powered by these solar panels.

34

: 00:02:11

And this is the kind of model where

today, if you're imagining a build

35

: 00:02:15

out of just Amazon data centers, you

wouldn't be able to even innovate to

36

: 00:02:18

say, what's a better way to build a

data center that might be different?

37

: 00:02:22

And they not only got a grant from

us, but after we kickstarted that

38

: 00:02:26

process, they also got a government

grant from the Netherlands.

39

: 00:02:31

Because it was so different from

what they were used to seeing.

40

: 00:02:34

And so I think that's the beauty of

decentralized physical infrastructure

41

: 00:02:38

is sometimes you can offer means to

experiment in new ways that today

42

: 00:02:44

nation monopolies may not necessarily

have as part of their own agenda.

43

: 00:02:48

And I think that's.

44

: 00:02:49

Something that's so exciting.

45

: 00:02:51

We also have definitely offered grants to

be able to, to verify, you know, outputs

46

: 00:02:56

coming from all of our data centers

at any point in time with dashboards

47

: 00:02:59

that show how much every single one

of our data centers that are on our

48

: 00:03:04

network, how much they're generating.

49

: 00:03:06

And I think sometimes this is.

50

: 00:03:07

Even better reporting than what we see

is ESG reports, which might be after the

51

: 00:03:11

fact reporting as opposed to something

that might be tracked in real time.

52

: 00:03:15

So there's so many different

use cases that people have done

53

: 00:03:19

from, again, like I said, the

infrastructure side all the way to.

54

: 00:03:23

Dashboard and reporting side, and then

also on what are the large climate change

55

: 00:03:27

data that today we wanna make sure are

coming from the right vetted sources.

56

: 00:03:32

Because often when you're looking

for information online, especially

57

: 00:03:35

if it's a polarizing topic where some

people may not believe the source

58

: 00:03:41

of certain data sets, we do have.

59

: 00:03:43

A great company called Equity Labs that

builds on top of the foco network that

60

: 00:03:47

has been looking at large amounts of

climate change data and making sure

61

: 00:03:52

that when we think about AI models to

the future, they're actually pulling

62

: 00:03:56

from the best sources out there.

63

: 00:03:58

So, so many angles to it.

64

: 00:04:00

We definitely care a lot about

climate and we have so many case

65

: 00:04:03

studies on our Follicle Green website.

66

: 00:04:05

You can just search Follicle Green

and you can see a handful of examples.

67

: 00:04:09

Anthony Perl: Amazing.

68

: 00:04:10

Thank you so much.

69

: 00:04:11

Well, I'm happy to jump in.

70

: 00:04:12

I just want to, I mean, I, I'm fascinated

a little bit about your background

71

: 00:04:15

and how you got to where you have,

because you've been on quite a journey.

72

: 00:04:19

So do you wanna fill us in

a little bit about that?

73

: 00:04:21

Clara Tsao: Yeah, I'm happy

to talk about my background.

74

: 00:04:24

I think for me, ever since I was

very young, I have always been

75

: 00:04:28

very, very passionate about my why.

76

: 00:04:30

I've always been very mission

oriented in spending my life working

77

: 00:04:35

on things that I really think

moved the needle in this world, and

78

: 00:04:38

that's what really gravitated me.

79

: 00:04:40

Towards not only working in startups and

also large companies, those gave me a lot

80

: 00:04:45

of learning on how, how the world moves,

how business deals are done to all the

81

: 00:04:50

way from working in the public sector.

82

: 00:04:52

So I spent a lot of years in the civic

tech community space, which mirrors

83

: 00:04:56

a lot of, you know, the community

activism we actually see in blockchain.

84

: 00:05:00

And there's a lot of people

that really believe that

85

: 00:05:04

technology should serve us all.

86

: 00:05:06

And what brought me into.

87

: 00:05:08

Working in the US government, and I

also advised 10 Downing Street, the

88

: 00:05:12

UK government in setting up a similar

innovation program, was this idea that

89

: 00:05:17

today most of the innovation we see in

this world are funded by Silicon Valley or

90

: 00:05:23

particular parts of the world that may not

actually care or have that incentive to

91

: 00:05:27

offer that technology every single corner.

92

: 00:05:30

And.

93

: 00:05:31

For me, being able to spend my time

working on technology that everyone

94

: 00:05:35

could access is more critical than ever.

95

: 00:05:38

And when you imagine building a product

in, in US government that everyone can

96

: 00:05:43

touch or has to touch in some form,

I see that as really impactful use

97

: 00:05:47

of technology and also being able to

help people access critical resources.

98

: 00:05:52

A lot of people in the US especially

fell behind when healthcare.gov

99

: 00:05:57

was a failure back in the day.

100

: 00:06:00

There was a SWAT team of very smart

technologists that all got together

101

: 00:06:03

and that same DNA of people was the

people that I worked with closely

102

: 00:06:08

when I joined an US government.

103

: 00:06:10

I always felt this strong sense of public

and civic service to really help as

104

: 00:06:15

many people as possible in my lifetime.

105

: 00:06:18

For me, the blockchain space is

especially impactful because a lot

106

: 00:06:22

of people are thinking in how we can

rethink the incentive structure of how.

107

: 00:06:28

Capital works in this world.

108

: 00:06:30

A lot of whether that's Bitcoin for

financial access in places like Argentina

109

: 00:06:36

where there might be large inflation

and people can't necessarily have

110

: 00:06:40

access to stability, and you see the

rise of stable coins as well, giving

111

: 00:06:45

people different alternatives to what

traditional financial systems are.

112

: 00:06:49

I think those are really important.

113

: 00:06:51

The the years I spent in issues around

online disinformation, I was the CTO

114

: 00:06:56

of two teams in the US government.

115

: 00:06:57

One focused on countering homegrown

extremism, so specifically how terrorists

116

: 00:07:01

are using the web to recruit and

radicalize, but also around election

117

: 00:07:06

security and foreign and influence

operations as this could be around how.

118

: 00:07:10

Different countries around the world

might try to manipulate elections

119

: 00:07:14

for other countries that might have

more of a geopolitical outcome.

120

: 00:07:18

Those are really hard and interesting

problems where I, I think really

121

: 00:07:22

understanding the architecture of the

web, how people communicate on the

122

: 00:07:25

web, that's more important than ever.

123

: 00:07:27

And so.

124

: 00:07:28

For me, my journey has

been actually very logical.

125

: 00:07:31

Every step of the way, I try to

find an even better way to solve a

126

: 00:07:35

problem, and I learn and cover more.

127

: 00:07:37

And sometimes that cross-disciplinary

path to where I am today has allowed

128

: 00:07:43

me to combine all these interests.

129

: 00:07:45

For example, today we are storing a

ton of government data sets on the Quin

130

: 00:07:50

network because back to my civic tech

roots, a lot of people don't realize that.

131

: 00:07:54

Data that even governments

release is, is fragile.

132

: 00:07:58

We've seen cities, mayors, governments

take down data from previous

133

: 00:08:03

administrations or previous regimes of

government because they don't wanna tackle

134

: 00:08:08

an issue that that data might expose.

135

: 00:08:11

Right?

136

: 00:08:11

So those things are so critical

for me, and coming back to the

137

: 00:08:16

internet infrastructure level

of how we can rethink storage.

138

: 00:08:19

Was a natural path when I saw all kinds

of content challenges when I was dealing

139

: 00:08:24

with the thorniest issues on the web.

140

: 00:08:26

The other part when I was working on

countering foreign influence operations

141

: 00:08:30

is you always wanna view attribute

where a source of data is coming from.

142

: 00:08:35

In 2016, there was obviously

a lot of media discussion and

143

: 00:08:40

in the UK saw it as well over.

144

: 00:08:42

Russia and a number of state sponsored

operations from Russia, uh, happening

145

: 00:08:47

in other elections around the world.

146

: 00:08:49

And that ability to say, Hey, this ad

actually came from a server farm in

147

: 00:08:55

Russia, was really, really important

for a number of people to really

148

: 00:08:59

realize, hey, these are not Americans.

149

: 00:09:01

On the other lines, making these

posts, those kinds of things have

150

: 00:09:05

severe outcomes to how people

participate in a democratic way.

151

: 00:09:09

All around the world to even

have a functioning democracy.

152

: 00:09:12

When people are unclear about the

information they can access and how they

153

: 00:09:17

can vote, that's very, very challenging.

154

: 00:09:19

And sometimes disinformation also prevents

great people from even being able to

155

: 00:09:24

be elected or run for office because.

156

: 00:09:27

There are a lot of rabbit holes

where the second you go down one

157

: 00:09:31

information silo, everything else

you're served is in that same bucket.

158

: 00:09:35

And at the end of the day, most companies

today that we see in the web, two version

159

: 00:09:40

of the web are built by an incentive model

where you have to be able to scrape user

160

: 00:09:45

data in order to sustain your business.

161

: 00:09:48

It's always an advertising play.

162

: 00:09:50

And this is how situations like Cambridge

Analytica took place where a lot of

163

: 00:09:55

people actually might be subtly aware

of their information being used, but

164

: 00:10:00

the large scale of how easily it was for

somebody to do that and monetize off of

165

: 00:10:05

user data and even, you know, use it for.

166

: 00:10:09

Particular research means against

their will was, was something

167

: 00:10:13

that a lot of people were shocked

by when it actually took place.

168

: 00:10:17

So I, I think these kinds of situations

were ones where, you know, for me,

169

: 00:10:22

being able to spend my time on.

170

: 00:10:24

On hard problems in this world is

a very strong motive for how I've

171

: 00:10:27

taken my career, and I hope to

continue going down that direction.

172

: 00:10:32

But that narrative has been very

consistent, even though I've worked on

173

: 00:10:35

different roles, every single one has

been a building block from the next.

174

: 00:10:39

And even the work I did in EdTech

around a decade or more ago was

175

: 00:10:44

really beautiful during the.

176

: 00:10:46

Pandemic COVID when kids had to remote

learn, and I never thought it would see

177

: 00:10:51

a day when there was an explosion of

ed tech everywhere because initially

178

: 00:10:55

you have to work with school systems.

179

: 00:10:57

There's all these regulation, and

when a pandemic hits like the one

180

: 00:11:02

we saw in 2020, it was amazing that

the infrastructure for ed tech was.

181

: 00:11:08

As mature as it had been, and it wouldn't

be there if there wasn't decades of people

182

: 00:11:12

really thinking a few years in the future.

183

: 00:11:14

And I think that's what's so

great about today, the blockchain

184

: 00:11:18

and crypto community is.

185

: 00:11:19

Sometimes it's hard to actually

tangibly feel the outcome, and we've

186

: 00:11:24

seen time over time more progression

to all kinds of tooling and innovation

187

: 00:11:30

that today, everyday consumers are

starting to understand more and more.

188

: 00:11:34

I think this year with

a lot of regulatory.

189

: 00:11:38

Upsides not only in stable coins,

but even, you know, in in the US

190

: 00:11:42

a lot of positivity in how people

understand crypto from the hill and

191

: 00:11:47

also regulators from around the world.

192

: 00:11:49

I think that's a big turn for

people to really feel like,

193

: 00:11:53

Hey, this is something that.

194

: 00:11:55

And its core technology has a lot

more good than what we might see on

195

: 00:11:59

the news for scams or failures from,

you know, people who are essentially

196

: 00:12:03

scammers at the end of the day.

197

: 00:12:06

Anthony Perl: What's the future though?

198

: 00:12:07

Where do you see it going?

199

: 00:12:08

Because you've seen this journey happening

and, and I think there's still a degree

200

: 00:12:12

of education to go around for people

to understand where the data's being

201

: 00:12:16

stored and what they're giving away.

202

: 00:12:19

But where's it taking us?

203

: 00:12:22

Clara Tsao: Yeah, I think where it's

taking us is, you know, ultimately I'm

204

: 00:12:26

super excited about the age of AI because

while there's a lot of challenges with

205

: 00:12:32

job replacement or which jobs will exist

and which ones won't, at the end of the

206

: 00:12:37

day, AI is making it easier for people

to find information more than ever.

207

: 00:12:42

Today, you don't even have to

be, for example, a programmer.

208

: 00:12:45

To learn how to code.

209

: 00:12:47

You can vibe, code and build

an application in seconds.

210

: 00:12:51

And so I think it's actually been

equalizer more than ever, as we see

211

: 00:12:55

at ai, actually augment and make

it easier to things more than ever.

212

: 00:12:59

My big look into the future is just

making sure that people are also

213

: 00:13:03

thinking about safeguards at the same

time, because with good there also

214

: 00:13:08

becomes bad actors that come into play.

215

: 00:13:10

And this is why I think the

best parts about crypto are.

216

: 00:13:14

Really the areas of people thinking

about verification and decentralization

217

: 00:13:19

because when one group gets access, too

much data, too much information, there

218

: 00:13:24

will always be ways that bad actors

take advantage of those systems for bad.

219

: 00:13:29

And we've also seen over the years

just significant it failures from

220

: 00:13:35

centralized systems like what happened in

CrowdStrike last year, or even Facebook

221

: 00:13:39

or Instagram going down, and just the

amount of people dependent today on.

222

: 00:13:43

Centralized infrastructure.

223

: 00:13:45

So I do think that the future

is going to be way more secure

224

: 00:13:50

in a more decentralized way.

225

: 00:13:52

And you're also gonna see a lot more

use cases that not just take place in

226

: 00:13:58

necessarily the defi side, but also

in areas like critical infrastructure.

227

: 00:14:02

When I was working in national

security, I did also oversee blockchain

228

: 00:14:06

technology and also IO OT adoption.

229

: 00:14:08

And a lot of what would be amazing if.

230

: 00:14:12

The tech was fully possible

is how we can use blockchain

231

: 00:14:15

to secure energy grids, right?

232

: 00:14:17

There's a lot of possibility in the

core tech that if we can actually

233

: 00:14:20

roll it out and go to market in

the right way, can really have a

234

: 00:14:24

lot of force for good for everyone.

235

: 00:14:26

And so that's my sense is, you know, the

future is brighter than ever, but it needs

236

: 00:14:31

to be decentralized and more people need

to understand core internet infrastructure

237

: 00:14:37

to be able to own all their data as

they're embracing the future with it.

238

: 00:14:42

Anthony Perl: Jemma, I'm interested

in your thoughts there about

239

: 00:14:44

the energy side of things.

240

: 00:14:46

Jemma Green: Well, I think for

decentralized storage, where

241

: 00:14:50

it's using like lazy assets or

repurposing old assets, that is.

242

: 00:14:56

Particularly interesting in terms of

not creating more infrastructure for ai.

243

: 00:15:02

I mean, if it is a lazy asset, it

would still be using energy, but

244

: 00:15:05

use less energy overall if it's a

lazy asset that can be put to use

245

: 00:15:09

using like the file coin ecosystem.

246

: 00:15:13

In terms of water, I mean, avoiding

building new data centers will

247

: 00:15:17

certainly reduce the amount of

water that's being consumed and also

248

: 00:15:21

the impact on local communities.

249

: 00:15:23

So I think these new models, new

infrastructure models, new commercial

250

: 00:15:28

models and business models that encourage,

you know, more circular economy and

251

: 00:15:32

repurposing of infrastructure is really

interesting 'cause the alternative.

252

: 00:15:37

Is, you know, some of the forecast

that you've seen for the amount of

253

: 00:15:41

energy that's needed, the amount

of data, new data centers that's

254

: 00:15:43

needed to meet, to feed the AI bs.

255

: 00:15:46

You know, some of it's un unbelievable.

256

: 00:15:48

And yeah, I think that the

experimentation that's happening within

257

: 00:15:54

the blockchain ecosystem, albeit.

258

: 00:15:57

And I think in your case, you know the

point you were making earlier, if I

259

: 00:16:00

understood correctly, is that if Defi is

the biggest use case, the addressable and

260

: 00:16:05

obtainable market for that is enormous.

261

: 00:16:07

Whereas you have a particular

service you wanna provide like.

262

: 00:16:12

Humanity as well as

general services as well.

263

: 00:16:16

And the attainable addressable

market for that looks different.

264

: 00:16:19

And so the idea that you should be

the same size as all the L ones are

265

: 00:16:23

looking at this, like the financial

markets as their use case is really

266

: 00:16:27

not a fair or appropriate comparison.

267

: 00:16:31

But in terms of like

data center growth for.

268

: 00:16:35

Let's say like the traditional means.

269

: 00:16:38

If it just goes down that path

and Web3 doesn't play a meaningful

270

: 00:16:42

part in that, I do feel like

it will be a bit of a failure.

271

: 00:16:47

I think all these other use cases are

really important and I don't wanna dismiss

272

: 00:16:50

them, but the, you know, the elephant in

the room is that right now, so maybe it's

273

: 00:16:55

not, file coin isn't the active, it's,

you know, that's not the project for you.

274

: 00:16:59

But do you see a role for that within

the Web3 ecosystem if it's not file coin?

275

: 00:17:04

And what are you seeing going on

around that space that you think is

276

: 00:17:07

particularly game changing and could

provide like a scalable alternative model?

277

: 00:17:12

Clara Tsao: Yeah, I mean I think like

Ashley, the entire deep end space and

278

: 00:17:16

it, it kind of got a lot of spotlight

in the last year or more is exactly

279

: 00:17:21

actually, like you said, much more

towards a future where we can think

280

: 00:17:25

about sustainable infrastructure, right?

281

: 00:17:27

You have the ability to not only

recycle data centers, repurpose

282

: 00:17:31

them in different ways, but also

think about decentralized compute.

283

: 00:17:36

Decentralized GPUs, decentralized storage,

all in one stack, rather than having to

284

: 00:17:40

build out brand new instances of that.

285

: 00:17:43

So we have a lot of storage

providers around the world that

286

: 00:17:47

have actually gone on to also work

with other DeepEnd projects to

287

: 00:17:52

offer different services rather

than just doing one size serves all.

288

: 00:17:57

I would say that at the end of

the day, um, I, I think it's

289

: 00:18:01

just different comparisons.

290

: 00:18:02

It's not saying good or

bad, but if you look at.

291

: 00:18:05

The way we measure success in the network

is actually if our goal is to store

292

: 00:18:11

and offer decentralized storage, the

actual Web3 market for types of data

293

: 00:18:17

to store is actually very, very small.

294

: 00:18:20

It's actually a tiny fraction of that.

295

: 00:18:22

So it's not that we.

296

: 00:18:24

We don't care.

297

: 00:18:25

It's just that we have a large set of

stakeholders we're always looking at,

298

: 00:18:29

and I think the upside of us leaning

into Web3 is really because a lot of

299

: 00:18:34

people are really good early adopters

that are willing to experiment and

300

: 00:18:39

willing to try and really willing to

innovate with us versus, you know,

301

: 00:18:43

some of the enterprise traditional

infrastructure providers might be a little

302

: 00:18:47

bit more skeptical about all kinds of.

303

: 00:18:50

Newer things that hit the market.

304

: 00:18:51

Right.

305

: 00:18:51

And that's, I, I wouldn't even

take it personally to Web3, that

306

: 00:18:54

that's all kinds of innovation that

are in the merging tech bucket.

307

: 00:18:58

But actually earlier this year, we were at

the World Economic Forum at Davos and the

308

: 00:19:02

idea of energy and, and data centers was.

309

: 00:19:06

Really important in this race

to win the AI race, right?

310

: 00:19:10

Like how do we do that?

311

: 00:19:11

For me, one of the best parts about

decentralized data centers and this idea

312

: 00:19:15

of not only storage but compute and GPUs,

is it really allows that flourishing of.

313

: 00:19:22

Wealth generation and job generation

to touch every single part of

314

: 00:19:27

the world, not just the few that

have resources or the talent.

315

: 00:19:31

Today, if we can offer, you

know, different entrepreneurs

316

: 00:19:34

in across Africa to to offer

data centers, people get better.

317

: 00:19:40

Local access to data, and they also

have the chance to also build that

318

: 00:19:46

economic pillar of opportunity.

319

: 00:19:48

So I think there's a lot

that is positive there.

320

: 00:19:51

I think unfortunately, there's

going to be energy consumption

321

: 00:19:56

in all forms, and I hope that.

322

: 00:19:59

At the end of the day, everyone thinks

about energy generation in the bigger

323

: 00:20:03

scheme of how it impacts everything

around them because there's future

324

: 00:20:08

generations that have to live on

this planet, that have to deal with

325

: 00:20:12

the long-term consequences of highly

energy, efficient on ways of doing so.

326

: 00:20:16

I do think that a lot of

people today, whether it's in

327

: 00:20:19

centralized A or decentralized

ai, are certainly thinking about.

328

: 00:20:23

More and more efficient ways

to lower the energy costs,

329

: 00:20:26

which I think is really great.

330

: 00:20:28

But the repurposing of critical

physical infrastructure, I think

331

: 00:20:31

is a huge part of this story.

332

: 00:20:34

Jemma Green: Mm-hmm.

333

: 00:20:35

And in terms of like GPU and compute,

is that something that is important

334

: 00:20:39

for Filecoin to offer services

around or is it mainly storage?

335

: 00:20:43

Clara Tsao: We actually are working

with a number of companies in both

336

: 00:20:47

decentralized GPUs and, and, uh, compute.

337

: 00:20:50

So Ather is a great example where we

have actually announced a partnership

338

: 00:20:54

with Ather many times to really

look at how we can help deploy

339

: 00:20:59

storage to their compute needs.

340

: 00:21:02

Mm-hmm.

341

: 00:21:02

And the same thing with Akash

for GPUs, we've worked with other

342

: 00:21:07

networks that offer other kinds of.

343

: 00:21:10

Computer or storage or GPU means to really

offer that trifecta of services for people

344

: 00:21:16

who have all kinds of storage needs.

345

: 00:21:19

That they wanna be able to not

only customize, but also do it

346

: 00:21:23

in a fully decentralized way.

347

: 00:21:25

So I think in the last year alone, we've

done probably 10 different partnerships

348

: 00:21:30

in the AI space, but that's all together

with other deep hip projects as well.

349

: 00:21:35

QP offer that entire AI stack because

you imagine the infrastructure layer

350

: 00:21:39

there is storage, compute, and GPUs.

351

: 00:21:42

And then on the training data layer,

there's a handful of companies really

352

: 00:21:46

thinking about how to have good training

data sets that are also decentralized,

353

: 00:21:51

that might also help creators monetize.

354

: 00:21:54

Off of their ip.

355

: 00:21:55

And then on the very top layer,

there's obviously people building

356

: 00:21:59

the application and AI agents that

are much more consumer facing.

357

: 00:22:02

And so I think there's so much that

is there that we're always constantly

358

: 00:22:07

trying to make sure we have the best

plumbing to be able to scale as we see

359

: 00:22:11

this whole entire ecosystem of other

players also also get into the spotlight.

360

: 00:22:17

Jemma Green: Great.

361

: 00:22:17

And in terms of like the ratio of

like storage, compute, and GPUs,

362

: 00:22:22

how would you break that up in

terms of all of your activities?

363

: 00:22:25

You mentioned like quite a lot of

partnerships that have been announced

364

: 00:22:28

that are more focused on the AI space.

365

: 00:22:30

Are they yet to become a really big

part of the overall data storage and

366

: 00:22:35

processing story, like in terms of volume?

367

: 00:22:38

Clara Tsao: Yeah, I think I said this

earlier, or maybe I thought it, but by

368

: 00:22:43

the end of 2026, 90% of new content on the

web are all gonna be AI generated, right?

369

: 00:22:50

And there has to be a way

to catch all of that data.

370

: 00:22:52

And I think that ai, whether

you are using it or not, is.

371

: 00:22:57

A huge chunk of where we're

gonna see data go for the future.

372

: 00:23:01

And so for us

373

: 00:23:02

Jemma Green: we are, but you, just to

interrupt you there you are talking

374

: 00:23:05

about storing content that was created

by AI as opposed to your storage system

375

: 00:23:11

being used to process AI queries.

376

: 00:23:14

Clara Tsao: Yes.

377

: 00:23:15

So there's the actual storage layer,

right, of how you store AI data.

378

: 00:23:20

There's also the store and, and storing

AI data can mean different forms, right?

379

: 00:23:23

It can mean like at the end of the

day, the content that AI is producing,

380

: 00:23:28

it can mean storing the training

data that's still stored somewhere.

381

: 00:23:32

Right, and that can also mean being

able to have the right infrastructure to

382

: 00:23:37

offer the storage, compute and GPU side.

383

: 00:23:42

And those are all important

ingredients that of course, that.

384

: 00:23:46

I think today, rather than having to

feel like you have one option down the

385

: 00:23:50

entire stack, you see a lot of players

offering that menu of optionality.

386

: 00:23:55

And I think that's a lot better in

the ecosystem like that than one where

387

: 00:24:00

you only have one choice or the other.

388

: 00:24:02

You know?

389

: 00:24:02

It's like you choose this, now

you have to go down the entire

390

: 00:24:06

stack this particular way.

391

: 00:24:08

Jemma Green: Oh, I see.

392

: 00:24:09

Yeah.

393

: 00:24:10

So it's kind of like the gateway as well.

394

: 00:24:12

Clara Tsao: Yes, exactly.

395

: 00:24:13

So Uhhuh.

396

: 00:24:15

Jemma Green: Yeah.

397

: 00:24:15

I appreciate you

distinguishing that for us.

398

: 00:24:17

'cause I hadn't really turned my

mind at all to storing all the, the

399

: 00:24:21

outputs of the AI data in addition to

obviously processing the things that

400

: 00:24:24

generate the output in the first place.

401

: 00:24:26

Yes, exactly.

402

: 00:24:28

Could you tell us your

favorite song at the moment?

403

: 00:24:32

Clara Tsao: Oh my God, I

have so many favorite songs.

404

: 00:24:36

I am about to watch a Coldplay

concert this weekend in Miami.

405

: 00:24:41

So probably the song

scientist is top of my mind.

406

: 00:24:45

I know Coldplay was in the internet

spotlight recently for other other means.

407

: 00:24:49

But

408

: 00:24:50

Anthony Perl: I was gonna say, talking

about AI using things to create lots of

409

: 00:24:54

images and content over the last few days.

410

: 00:24:57

Clara Tsao: Oh yeah, absolutely.

411

: 00:24:58

And it, it's funny that astronomer

actually, those inference data and

412

: 00:25:02

they actually got the most SEO hits

in their time when, when that scandal

413

: 00:25:08

and that incident broke on, on the

web with that Coldplay concert.

414

: 00:25:13

So it's kind of fascinating

to also see that spike.

415

: 00:25:16

But yeah, definitely a lot of people

have created names from that incident.

416

: 00:25:19

I'm sure everyone has,

has seen that video.

417

: 00:25:22

So.

418

: 00:25:23

Jemma Green: Yeah, it's um, a

lot of marketing for Coldplay.

419

: 00:25:25

Yeah, definitely.

420

: 00:25:28

Thank you so much for joining us.

421

: 00:25:30

It's been fascinating to do a, a dance

through like the history of the work

422

: 00:25:35

that you'd been doing and how that

led you to get involved with Filecoin.

423

: 00:25:39

And, you know, it's a topic that I'd say

most people don't know that much about.

424

: 00:25:43

And understanding some of the

new use cases that, and the

425

: 00:25:47

services that you're offering, as

well as alternatives for people.

426

: 00:25:51

And then an ecosystem where more

experimentation can happen so they're not

427

: 00:25:55

just having to use the cookie cutter that

the bigger providers have given them.

428

: 00:25:59

And, you know, without.

429

: 00:26:00

Any room to try new ways of doing

things, new commercial models,

430

: 00:26:04

new business models as well.

431

: 00:26:06

So I just wanted to say thank you

so much for coming on and helping

432

: 00:26:10

our listeners to understand more

about this very interesting and

433

: 00:26:13

important, relevant topic right now.

434

: 00:26:16

Clara Tsao: And thank you

so much for having me.

435

: 00:26:17

I'm always happy to share and

very, very also excited to hear

436

: 00:26:21

other episodes of your podcast

as well with all the incredible

437

: 00:26:24

guests that you have all the time.

438

: 00:26:26

So thank you so much.

439

: 00:26:27

Pleasure.

440

: 00:26:30

Anthony Perl: That's all for

this episode of Unblocked.

441

: 00:26:32

Please check out the show notes

for information on Power Ledge.

442

: 00:26:35

And other contact information.

443

: 00:26:37

We welcome your comments and

feedback and please hit subscribe

444

: 00:26:41

wherever you are listening.

445

: 00:26:42

This podcast was produced

by podcast Done for You.

446

: 00:26:46

We look forward to your

company next time on Unlocked.