0:00

Hey folks, Jeff Berman here. If

0:02

your business is driving innovation, delivering

0:04

exceptional experiences, or making real impact

0:06

on society, or maybe all three,

0:08

we want you to apply for

0:11

the Masters of Scale Business Awards.

0:13

These awards celebrate bold organizations of

0:15

all sizes and across all industries.

0:18

Award recipients don't just get a

0:20

trophy, although yes, there are trophies.

0:22

They get a spotlight at the

0:24

Masters of Scale Summit and a

0:26

seat at the table with the very best

0:29

in business. Don't wait. Head

0:31

to Masters of

0:34

scale.com/business awards dash

0:36

apply. Just separating

0:38

on what is real

0:40

from what is real from

0:43

what is hype I think

0:45

is very hard and I

0:47

don't think the AI industry

0:49

helps itself. with things like

0:51

not talking about hallucination rates

0:53

or not talking about what

0:55

does it take for something

0:58

to truly be enterprise grade.

1:00

There's a little bit of

1:02

a look much so cool

1:04

kind of attitude. I think

1:06

there is a maturity process

1:08

that is going to happen,

1:10

figuring out what is real from what

1:13

is hype is the biggest challenge

1:15

that business leaders face to

1:17

do. That's

1:26

Shridar Ramoswame, CEO of the

1:28

cloud storage platform Snowflake. It's

1:30

been a wild few years

1:32

for Snowflake, from a record-breaking

1:34

IPO to a plummeting stock

1:36

price, to a data breach

1:38

scandal. Shridar took over as

1:41

CEO in the heat of

1:43

the turmoil, and has helped

1:45

steady the ship in part

1:47

by leaning into AI in

1:49

a very specific way. Today

1:51

he shares lessons from the

1:53

company's in-process turnaround, including insights

1:55

behind high-profile high-profile partnership with

1:57

Open AI and

2:00

Antropic. Why snowflake?

2:02

China's deep-seek early,

2:05

and how the

2:07

buzz around agentic

2:10

AI is being

2:12

broadly misunderstood. There's

2:15

a lot to

2:17

cover, so let's

2:19

get to it.

2:21

I'm Bob Safian,

2:24

and this is

2:26

Rapid Response.

2:28

Stoneflake has been on quite

2:31

a roller coaster. It had

2:33

the biggest software IPO ever

2:35

in 2020, followed by a

2:37

somewhat dramatic stock fall. You

2:39

took over as CEO a

2:41

year ago and almost immediately

2:43

faced scrutiny related to a

2:45

high-tech data breach. Yet you've

2:47

also been in the heart

2:49

of like the hottest. business

2:51

arena ever. AI, Snowflake was

2:53

one of the first American

2:56

companies to tap into Deep

2:58

Seek, the China-based open source

3:00

AI. It's all kind of head

3:02

spinning. How's it going? Like has

3:04

your first year CEO been what

3:07

you expected? The first year

3:09

has been amazing. A little bit

3:11

more than what I expected

3:13

in terms of drama, but this

3:15

is a time of rapid change

3:17

and I could not be more

3:20

excited. I had a guest on the show

3:22

recently who confided that a lot of

3:24

CEOs are kind of paralyzed right

3:26

now by sort of external uncertainties

3:28

in the world, you know, shifting

3:30

tariffs and regulations and executive orders.

3:33

How do you deal with and

3:35

think about the sort of the

3:37

environment and all the changes relative to

3:39

sort of the things that you can

3:42

control yourself? One of my firm beliefs

3:44

in life is that you need to focus

3:46

on the things that you are going to

3:48

have an impact on. There are many

3:50

things that, let's face it, we are

3:53

simply not going to have any

3:55

impact on, obsessing about unchangeable

3:57

things in the short term.

3:59

the recipe for being uncertain about

4:02

life. There is a lot

4:04

of macro uncertainty. Businesses will

4:06

react and we will have to

4:08

worry. For example, if the stock

4:11

market keeps going down or if

4:13

the business climate gets worse, it

4:15

will have an impact on snowflake,

4:17

but so far it's been heads

4:20

down, get great product work done,

4:22

get great customer deployments done. We

4:24

first met when you were at

4:27

Google, advertising and commerce and then...

4:29

You started a search engine company,

4:31

Niva, that was ultimately acquired

4:33

by Snowflakes. Both of those businesses

4:35

were more content-focused and

4:38

ultimately consumer-oriented than Snowflakes.

4:40

What's different about a

4:43

full-on B-to-be business? There

4:45

are many things that are

4:47

similar in the sense that

4:50

you make money off of

4:52

deep relationships with your customers.

4:54

Absolutely, like commerce, search was

4:57

a consumer business. But Google

4:59

has an amazing enterprise business.

5:01

It's called ads. It predates

5:04

the Google cloud. And what

5:06

was unique about Google, of

5:08

course, was that a massive

5:10

amount of revenue on Google

5:13

ads is completely self-serve. It's

5:15

quite magical. 50% of this

5:17

absurd amount of revenue that

5:19

Google makes is without ever

5:21

talking to people. It's all

5:23

done via web interface. It

5:26

was quite magical. But the... core

5:28

of what I did at Google, which

5:30

is build highly reliable systems that never

5:32

went down. And for us, that was

5:34

all about money. If search ads went

5:37

down, you were literally losing thousands of

5:39

dollars every single second. That's enough

5:41

motivation to not bring stuff down.

5:43

And then work the relationships on

5:45

the enterprise side. There is a lot

5:48

of commonality to it, but obviously. It

5:50

is a new set of stakeholders. Google mostly

5:52

dealt with CMOs. Here we mostly deal

5:54

with CIOs and chief data officers and

5:57

thanks to AI quite a bit of

5:59

CEOs. as The

16:55

lobatical is just one of

16:57

the ways that local ensures their

17:00

employees feel appreciated and cared for.

17:02

And feeling appreciated is a principle

17:04

that is shared by their partnership

17:06

with Capital One business. We love

17:08

our 2% cash back card. We

17:10

can use the rewards to care

17:12

for our employees. My favorite thing

17:14

about Capital One, whenever I need

17:16

to call, there's always a caring

17:18

helpful voice on the other end.

17:20

You can't manufacture care, especially in

17:22

a big company, and Capital One

17:24

cares. the

18:10

state of Ohio. Whatever you're looking

18:12

forward to uniquely scale your

18:14

business, you can find it in

18:16

Ohio. Go to Jobs ohio.com to learn

18:18

more. Before

18:21

the break, Snowflake CEO Shrita Ramoswami

18:23

made the case that Snowflake is

18:25

the most consequential company in AI

18:27

and data in the world. Now

18:29

he shares what he learned from

18:32

last year's big data breach of

18:34

Snowflake customers. His use of weekly

18:36

warroom meetings and why AI hype

18:38

is the biggest challenge facing today's

18:40

business leaders. Plus, how he looks

18:42

at the immigration debate in the

18:45

US as an immigrant himself. Let's

18:47

jump back in. I

18:54

wanted to ask you about the

18:56

data breach controversy that hit last

18:58

year, soon after you came on

19:00

as CEO. A hundred and sixty-five

19:03

companies who used snowflake were impacted.

19:05

The breach came via customers' accounts,

19:07

not through your software or vulnerability,

19:09

but still, I'm sure the public

19:11

fracas wasn't welcome. I'm curious how

19:13

you faced that episode, whether it

19:15

was an advantage or a disadvantage

19:18

that you were early in your

19:20

tenure. And whether there were any

19:22

lessons you take away from it.

19:24

Security, we are very clear

19:26

with our customers. It's a

19:29

shared issue. Snowflake has offered

19:31

multi-factor authentication for over a

19:34

decade. We offered something called

19:36

network policies where he can

19:38

restrict who can connect to

19:41

your Snowflake instance. Having said

19:43

that, we are in it together

19:45

with our customers. Me going out

19:47

and saying that so and so is

19:49

that fault is simply not helpful. We

19:52

were very clear after doing a

19:54

lot of studies with external

19:56

partners that there were no

19:58

breaches or vulnerabilities in snowflake

20:00

systems. We also started putting in

20:02

place a series of schemes to

20:05

make sure that we would act

20:07

on our customers' behalf a lot

20:09

more quickly. We now have systems

20:11

that can detect access from surprising

20:14

places and then warn our customers,

20:16

or turn off accounts. We now

20:18

have dark web monitoring. So if

20:20

there looks to be a credential

20:22

that is compromised, then that credential

20:25

is promptly turned off. And, you

20:27

know, we used this opportunity also

20:29

to do other things, like make

20:31

sure that we had direct connections

20:33

to the security folks and all

20:36

our customers. So there was a

20:38

lot of learnings in terms of

20:40

how do we make sure that

20:42

we are in true collaboration with

20:45

our customers, how do we make

20:47

sure they understand that this is

20:49

a shared responsibility, and hopefully things

20:51

like the incident that we experienced

20:53

last year is a thing of

20:56

the past. When the crisis first

20:58

hit. I think a lot of

21:00

folks' impressions were, oh, this is

21:02

not going to be good for

21:05

Snowflake. And yet in some ways,

21:07

obviously you wouldn't want it to

21:09

happen again, but in some ways

21:11

it was good for your business.

21:13

Like, look, many of our customers

21:16

went through very unpleasant experiences as

21:18

a result of this. This is

21:20

not anything that I would wish

21:22

on anybody, but having had it

21:24

happen, you use it as an

21:27

opportunity to both make yourself the

21:29

relationships. and the overall posture much,

21:31

much better. You've attributed a lot

21:33

of snowflakes' recent progress to weekly

21:36

warroom meetings. I'm curious what those

21:38

meetings entail and whether we should

21:40

all be doing warrooms? Waterrooms, as

21:42

you know, they have particular connotations.

21:44

And one of my Google colleagues,

21:47

famously, I think it was 10-15

21:49

years ago, objected to waterrooms. So

21:51

I renamed one waterroom to be

21:53

a basket-weaving room. But... You know

21:55

the idea is how do you

21:58

bring together people quickly in effective

22:00

forums to help get past a

22:02

stumbling block that we have. Snowflakes

22:04

War Room was in the context

22:07

of the product engineering and marketing

22:09

and the sales teams needing to

22:11

work together, especially on new product

22:13

offerings. This is because Snowflake had

22:15

come of age as a cloud

22:18

data warehouse. We knew how to

22:20

sell the cloud data warehouse. We

22:22

didn't need to bring these different

22:24

functions together in order to figure

22:26

out how to sell these things.

22:29

On the other hand, something like

22:31

AI, we didn't know how to

22:33

position it. We didn't know what

22:35

our customers were exactly looking for.

22:38

We didn't know the kind of

22:40

problems that they were running into.

22:42

This is where the close collaboration

22:44

between the different teams that were

22:46

responsible for taking new products forward

22:49

were really helpful. And the warrooms

22:51

is much more in that context

22:53

of how do you do something

22:55

new that you know you're going

22:58

to struggle and get that to

23:00

a point of maturity. So it's

23:02

for a very very specific purpose.

23:04

Snowflake has invested in a lot

23:06

of AI startups. You recently announced

23:09

an expansion of your startup accelerator,

23:11

$200 million in new commitments. You

23:13

also announced plans to build a

23:15

new big AI hub at your

23:17

Menlo Park campus, a $20 million

23:20

AI upskilling program. How does this

23:22

fit together? What is this ladder

23:24

up to? Let's start with the

23:26

enablement investments. Our aspiration is to

23:29

train a million people. on using

23:31

AI and data products. And we

23:33

are doing this in a number

23:35

of different countries, including in places

23:37

like India, where an increasing number

23:40

of snowflakes customers, often based in

23:42

the US, are moving their technology

23:44

operations too. On the startup side,

23:46

we have a very healthy balance

23:48

sheet. It's over $5 billion. And

23:51

there are lots of startups that

23:53

want to build on top. of

23:55

Snowflake. Use Snowflake as a data

23:57

platform. So the $200 million fund

24:00

is in combination. with many venture

24:02

partners to fund the next generation

24:04

of companies. The AI hub is

24:06

more a physical space. We want to

24:08

host a set of companies that

24:10

want to experiment with AI.

24:12

And it's a continuation of

24:15

how we think about working

24:17

with the technology ecosystem to

24:19

help power the next generation

24:21

of companies. And we're just

24:23

happy to continue that because

24:26

we see that as being mutually

24:28

beneficial. businesses moving

24:30

to India. You are an

24:32

immigrant to the US from

24:34

India, you came from India

24:37

with just a few suitcases

24:39

and a couple hundred dollars,

24:41

as I recall. There's so

24:43

much angst in the US

24:45

around immigration right now.

24:48

How much do you think about

24:50

it, given your personal experience?

24:52

Look, I'm very, I'm

24:54

incredibly blessed. I

24:56

came with a bachelor's degree.

24:59

Yes, like I think it

25:01

was $700, $700. Neither of

25:03

my parents went to college.

25:05

I got a doctorate from

25:08

Brown that Brown entirely

25:10

paid for. I got

25:12

a monthly stipend and

25:14

a free PhD. And

25:16

I think I've contributed

25:18

in meaningful ways

25:21

to the country helping

25:23

create great, amazing

25:25

businesses. I think the larger

25:28

issue. is one that of

25:30

our population feeling like there is

25:32

enough prosperity to go on. People

25:35

in our country need to feel

25:37

like they have a prosperous future

25:39

before they're willing to lean in

25:41

and say we want more immigrants

25:43

to share in that prosperous future.

25:46

But I think those are the

25:48

core issues that our government needs

25:50

to address, where all of us

25:52

feel like they have the opportunity,

25:54

like I got the opportunity. My

25:57

take is there's no generosity without

25:59

prosperity. And as a

26:01

technologist, engaged in the creation

26:03

and the advance of technology,

26:05

it sure sounds like or

26:07

feels like it's enhancing opportunity

26:09

and prosperity, but that's not

26:11

translating to the broader public

26:14

necessarily. It's more than perception.

26:16

I think the honest question

26:18

that all of us need

26:20

to ask ourselves, not just

26:22

the immigrants, everybody, the government,

26:24

the population, is the prosperity.

26:26

truly broad? Are there truly

26:28

opportunities? Me arguing that, you

26:31

know, people like me are

26:33

going to create more companies

26:35

and more jobs if it

26:37

does not translate, you know,

26:39

call it prosperity in the

26:41

Midwest, is just not going

26:43

to resonate. You've said this,

26:45

you know, that you can

26:47

only unlock opportunity by embracing

26:50

change. And for you as

26:52

a business leader, trying to

26:54

embrace change. There's also risk

26:56

in moving too early when

26:58

things are changing so much.

27:00

How do you know what

27:02

change is sticky, especially when

27:04

headlines seem to blare about

27:07

new convulsions every day? Build

27:09

on strength. There are many

27:11

things that are very real

27:13

about snowflake. There is nothing

27:15

unreal about the three and

27:17

a half billion dollars that

27:19

we made as revenue, about

27:21

the 10,000 plus customers that

27:24

we have. about the mission-critical

27:26

role that we play. So

27:28

with AI, for example, we

27:30

didn't chase the hype. We

27:32

didn't say, oh, we'll fine-tune

27:34

and host models for you.

27:36

Here is a brand new

27:38

way of making money that

27:41

has nothing to do with

27:43

what Snowflake used to do

27:45

before. We said, let's offer

27:47

AI as a natural companion,

27:49

as a natural enhancer of

27:51

what you're already doing with

27:53

it. And so you need

27:55

to be very deliberate and

27:58

thoughtful about how you create

28:00

value, what your place in

28:02

the world is. And finally,

28:04

you also have to be

28:06

a little bit of a

28:08

portfolio manager. There are five

28:10

or six things that we

28:12

are going to try. are

28:15

not going to work out

28:17

and you have to have

28:19

unpleasant conversations, hard conversations internally

28:21

about what you need to

28:23

stop. We used to develop

28:25

foundation models but we backed

28:27

away from it because we

28:29

just said we just cannot

28:32

afford to spend the amount

28:34

of money and talent that

28:36

is required to train foundation

28:38

models. Was it an easy

28:40

conversation? Absolutely not. But I

28:42

think those are also important

28:44

conversations where you accept that

28:46

you're not going to know

28:49

everything. And when the outside

28:51

world speaks, you actually listen

28:53

and adjust and move along.

28:55

What are people and business

28:57

leaders most misunderstand about the

28:59

state of technology right now?

29:01

I think they are feeling

29:03

both pressure about things like

29:06

AI, but are also flooded

29:08

with options for what to

29:10

do. I think there's just

29:12

so much. just noise coming

29:14

in terms of partnerships between

29:16

X and Y or this

29:18

new agent take this or

29:20

the other. I think that

29:23

is just separating out what

29:25

is real from what is

29:27

hype I think is very

29:29

hard. I would say this

29:31

is less a misunderstanding than

29:33

an amount of confusion and

29:35

I don't think the AI

29:37

industry helps itself. with things

29:39

like not talking about hallucination

29:42

rates or not talking about

29:44

what does it take for

29:46

something to truly be enterprise

29:48

grade. There's a little bit

29:50

of a, look ma, it's

29:52

so cool, kind of attitude

29:54

to some of the things

29:56

that happen in AI. I

29:59

think there is a maturity

30:01

process that is going to

30:03

happen, but I think figuring

30:05

out what is real from

30:07

what is hype is the

30:09

biggest challenge that business leaders,

30:11

enterprise leaders face today. Shudard,

30:13

this was great. Thanks so

30:16

much for doing it. Thank

30:18

you, Bob. Really appreciate chat.

30:20

Shruder may be soft-spoken, but

30:22

he doesn't mince his words,

30:24

especially when it comes to

30:26

AI. I agree that separating

30:28

what's real from the AI

30:30

hype is critical for today's

30:33

business leaders. And if you're

30:35

a non-technical business leader, that

30:37

challenge is even harder. As

30:39

for Shruder's own hype, is

30:41

Snowflake actually the world's most

30:43

consequential company in AI? Might

30:45

if it come that one

30:47

day, I don't know. Still,

30:50

I feel smarter after listening

30:52

to him. There's a thread

30:54

that runs through Shriters' comments

30:56

that resonates from me beyond

30:58

AI about confusion in the

31:00

face of a lot of

31:02

noise. Whether we're talking technology,

31:04

tariffs, immigration, or whatever else,

31:07

we strive to anchor on

31:09

what's solid. Some of our

31:11

confusion is because we want

31:13

clear answers when there just

31:15

aren't any. And even the

31:17

most advanced AI won't solve

31:19

that problem. That's just something

31:21

we have to learn to

31:24

live with. I'm Bob Safian.

31:26

Thanks for listening. But she

31:28

said, you know, let's do

31:30

it. Let's just start our

31:32

own brand from scratch. Romeo's

31:34

recalling the moment when he

31:36

and his wife and co-founder

31:38

Milca Regali decided to take

31:41

a leap of faith. I

31:43

started working as a server

31:45

at Milca's mom's restaurant. I

31:47

fell in love so much

31:49

with the industry, and that's

31:51

what sparked it. Romeo and

31:53

Milca weren't certain how they

31:55

would bring their dream to

31:58

fruition, but they were certain

32:00

of one thing, their passion.

32:02

We knew we had a

32:04

vision and we found a

32:06

space. We had to gut

32:08

the entire... and build everything

32:10

from scratch. The kitchen, gas

32:12

piping, and the restroom, the

32:14

sound system, everything. We really

32:17

believed every detail matters. As

32:19

they broke ground on their

32:21

first raw's location, Romeo and

32:23

Milka soon faced the financial

32:25

reality of building something from

32:27

scratch. They looked to Capital

32:29

One business to help navigate

32:31

the fiscal burden of making

32:34

their dreams come true. We

32:36

used a spark cash plus

32:38

card from Capital One. The

32:40

no-pre-set. spending limit really had

32:42

a big role in helping

32:44

us finish the project. We're

32:46

very happy with what we

32:48

have accomplished. We want to

32:51

expand more. To learn more,

32:53

go to capital1.com/business cards. AI's

32:55

impact on the environment is

32:57

one of the most pressing

32:59

issues facing the tech industry

33:01

today. People want to know.

33:03

What's the carbon footprint of

33:05

a chat chipeti query? What

33:08

does it mean to innovate

33:10

sustainably? And can AI actually

33:12

be used to solve the

33:14

climate crisis? I'm Rana El

33:16

Caliwi. On my podcast, Pioneers

33:18

of AI, we bring questions

33:20

like this to some of

33:22

the leading thinkers and builders

33:25

working in AI. Join me

33:27

each week as we explore

33:29

how this technology is leaving

33:31

its mark on humanity and

33:33

our planet. Find pioneers of

33:35

AI on Apple podcasts, Spotify,

33:37

YouTube, or wherever you get

33:39

in your podcasts. Rapid Response

33:42

is a wait-what original. I'm

33:44

Bob Safian. Our executive producer

33:46

is Eve Tro. Our producer

33:48

is Alex Morris. Associate producer

33:50

is Mashamaku Tonina. Mixing and

33:52

Mastering by Aaron Bastinelli. Our

33:54

theme music is by Ryan

33:56

Holliday. Our head of podcast

33:59

is Lyttal Malad. For more

34:01

visit... Rapid Response show.com