The next frontier in cloud

computing: Decentralization
as a catalyst for disruption
July 2024

Mahesh Ramakrishnan
Escape Velocity

The story of David vs. Goliath resonates deeply, epitomizing a recurring motif in human
history: the unexpected triumph of smaller, innovative forces over formidable giants.
This narrative is strikingly relevant in the history of corporate strategy, where the winds of
creative economic disruption have consistently overturned established regimes and
reshaped the competitive terrain. In cloud services, we are potentially witnessing the
Christopher Jensen emergence of another pivotal confrontation.
Director of Research
Franklin Templeton Digital Assets
The cloud services industry boasts a small number of extremely powerful scale econo-
mies, as Telecom Giants (AT&T, T-Mobile) and Web2 incumbents (Amazon, Microsoft)
have created infrastructure monopolies, taking advantage of a historically low cost of
capital and robust government subsidies. These oligopolies can use their infrastructure
advantages to king-make companies in areas of emerging technology, the strength of
which was exemplified by Microsoft’s investment in OpenAI, a generational technological
asset. In a break from traditional investments and mergers, OpenAI accepted $10 billion
Co-sponsor
Prof. Jorge Tamayo of compute credits and infrastructure support from Microsoft in return for 49% of the
Harvard Business School company, an unprecedented deal both in terms of size and structure.1 How does one even
begin to compete?

Network economies offer a potential solution, as they derive a strategic advantage

from decentralized collaboration as opposed to centralized scale. Crypto-powered
infrastructure is exemplified by projects such as Filecoin that can generate the necessary
network effects to credibly compete with incumbents by turning people and small
businesses into infrastructure suppliers. By aligning upfront capex, labor, and rent costs
 across network participants, decentralized physical infrastructure (“DePIN”) networks can
create digital commodities like storage and bandwidth cheaply. A prime example, the
Filecoin network powers verification, storage, and retrieval for third parties who want to
provide cloud services to businesses. Its disruptive business model facilitates the creation
of a competitive storage marketplace financed through the fees that decentralized
storage providers (SPs) earn for the storage and retrieval of data and incentivizing
suppliers with its native currency ($FIL). By joining the Filecoin network, businesses with
excess storage capacity have a chance of competing with Amazon’s centralized scale
on cost by becoming SPs, with high guarantees around security and access. Today, DePIN
networks might be the single biggest threat to AWS/Azure’s cloud duopoly.

But before we can analyze the opportunity for disruption, we must first understand what
has driven incumbents’ competitive advantages. In short, it’s the combination of scale
advantages, regulatory barriers, and strategic market control—these incumbents also
influence downstream players reliant on their rails and use their scale to stifle competition.
The high barriers to entry for new market participants are considerable, characterized
by high capital costs, entrenched supplier relationships, and steep startup costs. However,
as traditional monetization models face growing challenges, the opportunity for chal-
lengers with better unit economics is growing.

Economic change has weakened the business model of incumbents. Traditionally, cloud
service firms are optimized through software-as-a-service (SAAS) frameworks allowing
high utilization levels on their modular architecture. These “one size fits all” subscription
economics, however, don’t always make the most sense for the end customer. Driven by
increasing data intensity of products, clients are pushing the market towards a pay-as-
you-go model for digital infrastructure and providers have had to respond by evolving the
SAAS model to one driven more by metered usage, termed infrastructure-as-a-service
(IAAS). 2 The IAAS model, while convenient for the customer, reintroduces uncertainty that
complicates unit economics by removing the key fourth step of the SAAS process (illus-
trated below). Higher uncertainty inevitably leads to price competition and a need to
recoup outlays through volumes quickly. Margins are driven lower by this, and growth is
stifled as uncertainty increases financing frictions. A new way to monetize customers
requires a new strategy, with a refocused plan around marketing, sales, etc. The compar-
ison between these business models is laid out below in Exhibit 1.

Exhibit 1: Business SAAS IAAS

Model Shift Acquire Hardware Acquire Hardware
Use Supplier Power to Drive Discount Use Supplier Power to Drive Discount
capacity to acquire hardware
Leverage frees up more

Lease Real Estate Lease Real Estate

To Build the Datacenter To Build the Datacenter

Sell Long-Term Subscriptions Pay-As-You-Go Model

2–3 Years in Term Paid Monthly Uncertainty in Payment Term and Churn

Leverage Up Scale At Pace of Payback

Use Long-term Sub Cash Flows to Support Debt Use Cash Flows to Support Buying New Hardware
Repayment with High Certainty

For illustration purposes only.

2 The next frontier in cloud computing: Decentralization as a catalyst for disruption

 Rather than rely on predictable contracts, IAAS providers must forecast demand. This
process introduces assumptions that increase uncertainty of cash flows, making debt
financing more difficult than before. This change forces incumbents to price services on a
variable unit basis, rather than a bundled fixed basis. This variable unit price is far more
subject to market competition, as the lack of long-term contracts reduces switching costs
and increases incentives for competitors to use promotions, further affecting customer
loyalty and margins. Inevitably, markets defined by commoditized services compete on
price—high fixed costs and high uncertainty usually point to the use of meaningful promo-
tion to capture attention. Recouping upfront cost at any price becomes the focus as
management teams become minutely focused on financial performance—nowhere is this
paradigm clearer than in the airline industry, where Boeing and Airbus discount massively
to keep a high level of turnover. While the need for this new strategy is driven by changing
demand-side needs, it’s also driven by advancements that increase supply side availability.
Summarized in three factors below in Exhibit 2, these changes have seeded the ground for
alternatives to proliferate.

Exhibit 2: Market
Considerations 1 2 3
People are in possession of more The size of the internet has Composable software powered
powerful hardware, and software 100×’d, leading to an explosion in by blockchain allows trustless
has become an increasingly the amount of data available, and collaboration among parties that
crucial driver of infrastructure computation to make stored data may not work together otherwise
efficiency and service quality more valuable

Cryptoeconomic networks can offer better cost, security,3 and product quality than
incumbents. While Bitcoin, the earliest cryptoeconomic network, organized big groups of
machines to collaborate, it didn’t focus on output beyond hashing—thus while Bitcoin
controls one of the biggest supplies of compute in the world, that computer is not opti-
mized to be economically useful other than validating the network to process payments.
The next network to reach scale, Ethereum, incentivized the creation of useful services,
allowing developers to access compute power they could source from peers to build
apps. Filecoin has gone a step further, creating a blockchain-based protocol geared
toward making it easy for anybody to sell storage services to clients, with a transaction fee
model ideal for constant-use IAAS use cases. Moreover, because Filecoin is an open-
source, distributed network, hundreds of developers have built tools and services that
make provision of storage services to clients extremely easy for businesses with excess
resources. Filecoin incentivizes SPs to participate in the network: SPs earn the blockchain’s
native coin, ($FIL), for doing business on the Filecoin network, further accelerating SP
economics and driving down service price. Without a similar scale of upfront infrastructure
costs, and being able to use Filecoin’s payment network, these SPs can provide storage
services just like AWS at a far cheaper cost.

Mining—i.e., joining the network as a SP—is largely carried out by businesses with excess
capacity, and groups of individuals co-locating hardware: There are ~2,800 distinct SP
systems on the network today providing storage services. Some of the biggest SPs on the
network include Seal Storage, 1475, Holon, DCENT, and Greater Heat, who have onboarded
significant network capacity and serve multiple enterprise clients. Seal, as an example,
is focused on onboarding high-density academic research and has historically optimized
their platform around long-term storage of that data. Holon focuses on cutting-edge,

3 The next frontier in cloud computing: Decentralization as a catalyst for disruption

Exhibit 3: Filecoin
Platform User Storage User
Device Provider Device

1 2 3 4
User pays storage Storage provider The network constatly User pays storage
provider to store commits publicly via verifies that storage provider to retrieve
their file Filecoin’s blockchain providers are storing their file back again
to storing the file files correctly

Source: Reidhead, Kyle. “Is Decentralized Cloud Storage the Next Big Web3 Industry?.” Web3 Academy. February 20, 2023.4

eco-friendly storage mechanisms combined with edge computing. Each SP can decide
what tradeoffs they support, the specific nature of the service they intend to provide and
build a custom service delivery plan around it. They transact mainly in $FIL and receive an
incentive for doing business on the network as well as payment from the client. While
enterprise-grade SPs may build out a whole support infrastructure, many of the SP busi-
nesses are fewer than 20 people and are taking advantage of excess hardware they
already own as part of a bigger business.

Organizations across the Filecoin ecosystem have also invested significant resources over
the last two years to dramatically increase Filecoin’s usability—an essential component
for Filecoin to become truly enterprise-grade and enterprise-ready. With this influx of
network orchestration capabilities, the network is primed for more sellers and users as the
network’s value is more easily realized and demonstrated. Since November 2023, the
ecosystem has welcomed multiple new data onboarding software platforms into the
community with more still planned to go live later this year. In response, the proportion of
big datasets stored by SPs has rapidly increased, with only ~50% of capacity being
used for small projects (0–10 terabytes) today vs. 90% two years ago. This suggests SPs
are attracting higher-quality institutional clients as Filecoin’s UX improves. How the
Filecoin model creates a ripe environment for SP businesses to drive value is laid out
below in Exhibit 4.

Exhibit 4: Strategy
Considerations 1 2 3 4
Storage providers with Limited opportunity cost Filecoin banner allows Scale of Filecoin further
hardware can’t compete to join the Filecoin SPs to weird purchasing reduces barriers to new
with AWS/Azure pricing network, offering power of a scaled player SPs joining the network
at scale, leading to immediate monetization when negotiating deals in a virtuous cycle
excess capacity of excess capacity

4 The next frontier in cloud computing: Decentralization as a catalyst for disruption

 Filecoin is autonomous, and its token incentive system regulates and encourages appro-
priate behavior from SPs and users. Economic activity is governed by the proof of
“spacetime” consensus algorithm, which empowers validators to continuously verify that
stored data is secure and hasn’t been changed, rewarding them with the Filecoin token
for their contributions. Notably, SPs must stake Filecoin tokens in concert with their
provision of storage services: bad acting SPs have their stakes slashed in response to
uneconomical behavior, therefore ensuring economic activity on the network is produc-
tive. The full storage and retrieval process is laid out below in Exhibit 5.

Exhibit 5: Filecoin Miners Bid

Data is Fragmented,
Data Retrieval
Encrypted and Stored
Process Diagram to Store Data Request
Client Given the
Only Existing Filekey

Client Sends FIL

to the Quickest
Mining Power Requires Retrieval Miner
Performance Hardware
HDDs, CPUs, GPUs,SSDs,
RDIMMs, Motherboards
Storage Power Requires
Data is Recorded
Sealing Storage
in the Blockchain
RAM, High-Density HDDs,
(IFPS)
SDDs, and future mining
Storage miners are hardware
Data Input paid in installments
Request
with FIL

Data Retrieval Requires

Powerful Processor
GPUs, CPUs and future
mining hardware Retrieval Miner
Retrieves Data

Source: “The Release of Filecoin Could Mean Product Shortages in the Semiconductor Industry.” Fusion Worldwide. December 11, 2020.5

The Filecoin network is extremely scalable and doesn’t bear any of the operating costs of
the SPs themselves. It offers an avenue by which SPs are brought together on a common
platform, allowing them to benefit from network effects and offer products that compete
with Amazon on the price they charge for a unit (GB) of storage. Most big, centralized SPs
like AWS will store a set number of copies of your data on their closed systems (usually
three copies). But on the Filecoin network, users can choose to store many copies of their
data—in whole or in part—encrypted or non-encrypted—across many nodes on the
network. Therefore, if one node fails, the data persists.

Data verifiability differentiates Filecoin from other centralized SPs whose solutions present
more like a black box. When dealing with confidential data, provenance is increasingly
important. Because all information is transparently stored in highly secure places that
clients can verify themselves, Filecoin is well suited for high-sensitivity datasets that
can’t afford data loss/modification. According to DeStor’s Filecoin Network Data Client
Explorer,6 major storage clients range from The Starling Lab and the Internet Archive to the
Web3 platform OpenSea and the Layer-1 network Solana, which stores its entire block
history on Filecoin. We also see it commonly used to store large research datasets for
universities, as well as highly sensitive medical and genetic record data due to the security

5 The next frontier in cloud computing: Decentralization as a catalyst for disruption

 guarantees provided by verifiability. Solana, which considers Web2 giants among its
biggest competitors, potentially has issues storing its entire history on AWS where
a centralized competitor could evaporate its entire business with a misclick. Filecoin
eliminates this platform storage risk.

Filecoin’s fee-based transaction model also makes it uniquely suited to serve pay-as-
you-go IAAS use cases, which are better aligned to its business model than that of
centralized giants. Today Filecoin generates fees7 (~$23 million in 2023 at the current token
price)8 from clients to SPs. These clients are typically paying for storage in $FIL directly,
but also have the capability to pay in stablecoins. SPs and storage clients negotiate
deals for duration, amount, and price directly. Fees can be inherently seasonal for large
clients who store big troves of data where they typically pay storage and transaction
fees upfront and annually.

The majority of Filecoin’s costs are paid through token emissions, meaning Filecoin mints
new tokens to distribute. This also means that most costs aren’t paid for in cash and
that protocol revenue is actually used to buy and burn tokens thereby reducing net new
emissions, a mechanism similar to a buyback. Filecoin’s structure follows Chris Dixon’s
model for Overcoming the Bootstrapping Problem:9 Token holders bear the costs of
Filecoin’s growth as it scales, expecting that ecosystem growth will result in bustling
economic activity that makes the protocol valuable.

Below (Exhibit 6) we show how Filecoin’s cash flow generation would change if the
revenue the protocol generated 50×’d overnight to illustrate how scalable the model is.
If revenue 50×’d, miner incentivization costs would remain the same (the subsidy isn’t
scaled as a function of revenue as it’s programmatically encoded), as would Filecoin’s
R&D costs. The only cost that would increase would be its sales and marketing costs
from additional headcount. As such, as Filecoin significantly increases revenue, it also
swings to being cash flow positive. Note that because Filecoin conducts economic activity
in their $FIL token, this analysis is priced in millions of $FIL to elucidate some of the
mechanics at play and intends to show how viral adoption of the new products coming to
market might affect unit economics. Simply put, new demand on Filecoin generates
cash flow at a rapid rate.

Conducting the analysis in Exhibit 6 in $FIL terms illustrates how much operating leverage
there is within the Filecoin system. The Filecoin network is just software—there isn’t a
10,000-person team somewhere managing this—there is no central team at the core.

Exhibit 6 (Millions of Filecoin Today ($FIL) Filecoin 50x Today ($FIL)

$FIL): Filecoin Scale Revenue 4.9 245.0
Gross revenues in $FIL scale without
– Labor Cost of Running Data Centers 0.0 0.0
any change in cost or incentives
– Miner Incentivization Costs (49.7) (49.7)

Gross Revenue (44.8) 195.3

– Research & Development (10.7) Only USD cash expenses (sales & (10.7)
marketing) rise as revenue explodes
– Sales & Marketing (0.7) (36.8)
– General and Administrative (3.6) (3.6)

Illustrative EBITDA ($FIL) (59.7) 144.4

EBITDA Margin NM 59%

Sources: Company Filings, Messari, EV3 and Franklin Templeton Analysis.

10

6 The next frontier in cloud computing: Decentralization as a catalyst for disruption

Exhibit 7: Filecoin Filecoin EBITDA Margin Scenario Analysis
Sensitivity Sales and Marketing Margin

5% 15% 25% 35%

1 NM NM NM NM
20 30% 20% 10% NM
Multiple of 50 69% 59% 49% 39%
Network Size
100 82% 72% 62% 52%
vs. Today
by Revenue 500 92% 82% 72% 62%
1,000 94% 84% 74% 64%
10,000 95% 85% 75% 65%

Sources: EV3 and Franklin Templeton Analysis.

Other than marketing, costs don’t really increase even as revenues scale 50 times—EBITDA
margin swings as high as 59% with this increase and could go as high as 85% as revenues
scale further. Most costs today are incentives to help SPs compete on the network but
should the network scale massively, one would imagine you’d need fewer incentives to
keep suppliers engaged. As the network scales revenue by bringing productive storage
onto the network, the fees earned directly remove $FIL tokens from circulation and deliver
greater ownership to token holders.

There is a reflexivity piece to consider here as well: because Filecoin’s costs are paid in
$FIL tokens, an increasing token price can increase costs to the network. The same
number of tokens is paid to SPs, but the dollar value of those tokens is higher. Below
(Exhibit 8) we present an illustrative analysis of what the cash flow generated by the

Exhibit 8 (Millions Filecoin 2023 50× Pre-Buyback ($MM) Filecoin 50× ($MM) Post-Buyback
of Dollars): Revenue 1,102.5 1,102.5
Conservatively assume revenue
Filecoin Buyback – Labor Cost of Running Data Centers 0.0 doesn’t increase with token price 0.0
– Miner Incentivization Costs (223.7) (447.3)

Gross Revenue $878.9 $655.2

– Research & Development (47.9) Reflexive analysis assumes removing (95.9)
7% of circulating supply doubles
– Sales & Marketing (198.5) (198.5)
token price
– General and Administrative (16.0) (32.0)

EBIT $616.5 $329.0

Filecoin can scale revenue massively
– Interest Expense 0.0 without scaling costs at all 0.0

Operating Income $616.5 $329.0

+ Depreciation and Amortization 0.0 Assumes 2023 EBITDA is used to buy 0.0
back $FIL tokens
+ Interest 0.0 0.0
– Capex 0.0 0.0

Illustrative EBITDA $616.5 $329.0

+ Non-Cash Costs Paid With Token Dilution 287.6 575.1
Illustrative Cash FCF $904.1 Rising token price increases $904.1
accounting costs, but doesn’t affect
EBITDA Margin 56% 30%
cash generation
Cash FCF Margin 82% 82%

Illustrative analysis does not consider taxes and assumes the buyback of tokens has caused price to double to $9/FIL
Source: EV3 and Franklin Templeton Analysis.11

7 The next frontier in cloud computing: Decentralization as a catalyst for disruption

 network would look like in US dollar terms, considering the tokens removed from circula-
tion as part of the buyback. We look at the effect of using the fees generated from
the course of business to remove 7% of the total tokens at a fixed price and assume that
the token price doubles in response to the supply shock. Because of the reflexive
token structure, a higher token price means Filecoin is paying out more in incentives (in US
dollar terms), which has the effect of bringing more SPs into the ecosystem. However,
while the network looks less profitable on the face, it still generates just as much cash flow
because while token issuance is dilutive, it isn’t a cash cost. We also conservatively
assume revenue is unchanged, when a higher token price and increased usage would
likely increase revenue in US dollar terms as well.

Using this analysis, we can clearly see that while reflexivity leads to higher headline costs
for the Filecoin network, those costs are non-cash costs that don’t affect the cash genera-
tion of the protocol. Free cash flow is still ~$900 million in either case.

Next in Exhibit 9, we compare what a scaled Filecoin would look like compared to a Web2
competitor like AWS in US dollar terms. Note this isn’t a perfect comparison (comparing
a company vs a protocol), as AWS has hardware expenses and a full organization behind it,
and Filecoin’s incentives are a non-cash expense. The most important difference to
call out immediately here is the difference in fixed costs: while Amazon has signicant fixed
costs, Filecoin has none. The largest difference is capital expenditures: while Amazon
spent $22 billion on hardware capex to create and scale datacenters and another
$22 billion on operating costs of those datacenters (real estate, energy), Filecoin has no
such comparable costs. Filecoin covers this by directly incentivizing SPs, who themselves
spend the capex to buy hardware with storage capacity and keep the machines running.

Exhibit 9 (Millions of AWS 2023 ($MM) Filecoin 50× ($MM) Post-Buyback

Dollars): Filecoin vs. AWS Revenue 90,757.0 1,102.5
Filecoin has no labor cost, variable
– Labor Cost of Running Data Centers (5,445.4) cost is paid in token inflation to miners. 0.0
– Miner Incentivization Costs (21,781.7) AWS bears this alone (447.3)

Gross Revenue $63,529.9 $655.2

$FIL emissions drive G&A and R&D
– Research & Development (11,671.8) spend. There is no organization to (95.9)
– Sales & Marketing (16,336.3) support vs. AWS which has >130,000 (198.5)
employees
– General and Administrative (10,890.8) (32.0)

EBIT $24,631.0 $329.0

FIlecoin can scale revenue massively
– Interest Expense (2,713.3) without scaling costs at all 0.0

Operating Income $21,917.7 $329.0

+ Depreciation and Amortization 21,706.6 Mining drives capex to the edge of 0.0
the network, and is accounted for in
+ Interest 2,713.3 0.0
infrastructure cost
– Capex (24,504.4) 0.0

Illustrative EBITDA $21,833.2 $329.0

+ Non-Cash Costs Paid With Token Dilution 0.0 575.1
Filecoin costs are paid in tokens—
Illustrative Cash FCF $21,833.2 $904.1
the network uses limited cash other
EBITDA Margin 24% than sales and marketing 30%
Cash FCF Margin 24% 82%

Illustrative analysis does not consider taxes and assumes the buyback of tokens has caused price to double to $9/FIL
Source: EV3 and Franklin Templeton Analysis.

8 The next frontier in cloud computing: Decentralization as a catalyst for disruption

 Filecoin pays this cost in tokens (created through dilution) but doesn’t have to put up any
physical infrastructure themselves. By serving as a platform for other storage providers,
Filecoin remains capital light even as they provide a capital-intensive product.

In Exhibit 9 on the previous page, we allocate token emissions to Filecoin cost buckets as
mapped by the whitepaper and show illustrative EBITDA on a comparable business to
what Amazon lays out for AWS in its filings. To take a snapshot here, we assume an initial
fixed price of $4.5 per Filecoin ($FIL) token that doubles because of the buyback. Finally,
we illustratively assume, like we did above, that any sales/marketing expense Filecoin
bears is paid in cash.

Once Filecoin is at scale, as fixed token emissions don’t really change, revenue scales
massively. As such, Filecoin can produce higher margins at scale than Amazon, and
explosively so in cash terms. This analysis also hits home the value of Filecoin’s incentive
structure on decentralizing capex, a massive contributor to cash burn for Web2 busi-
nesses: the network benefits from scaling usage without scaling fixed costs. Overall,
Filecoin’s structure at scale is significantly more flexible and profitable than AWS, which
suggests that should it get to scale, it could be a runaway winner.

While the potential for Filecoin to dominate the cloud services markets clearly exists, one
must question why it hasn’t happened: after all, Filecoin has been around since 2014. Part
of the answer relates to technological progress. The largest technical challenge in
Filecoin’s early days related to latency: consensus mechanisms require high download
speeds and low latency to coordinate across geographically dispersed machines.
Without high speeds, the performance of Filecoin significantly lacks central providers with
private networks as slower consensus drives slow service speeds around storage and
retrieval. Filecoin requires today’s mobile infrastructure (which have 10×’d data speeds in
the last decade12) as a baseline to enable performance in line with AWS. Similarly, the
nature of computation capable devices has rapidly accelerated: an iPhone today has
transistor density 171,000 times (!) greater than a supercomputer in 1980, and 90× greater
than an Apple Macbook Pro in 2010. The proliferation of compute-capable devices has
increased the pool from which Filecoin can pull storage capacity, resulting in higher
quality capacity, and better service (more available devices to store results in faster
retrieval and proximate caching).

However, even with these improvements, worryingly, utilization of the network remains
low at ~18%. We suspect three causes: a combination of initial oversupply of capacity,
tough business development tradeoff choices going after cold storage which drove
a UX problem with data retrieval being a clunky process, and the fact that we’re super early
and compute-over-data is just taking off, as we highlight below. On the demand side,
protocol revenue has also been extremely volatile due to early storage products on
Filecoin not finding product market fit themselves. While storage is a commodity (like oil),
how it is delivered varies depending on the data type and end use in computation
(similar to how there are multiple ways to refine oil). With an initial gearing toward cold
storage, the teams building on the network underinvested in easy-to-use retrieval for
the data they stored. This led to a clunky user experience, where people would find it easy
to store data, but hard to retrieve what they needed instantaneously. As such, early
clients were limited to those focused more on provenance and less on accessibility.

9 The next frontier in cloud computing: Decentralization as a catalyst for disruption

 This is rapidly changing as a few companies are trying to tackle scaling retrieval including
Banyan and Flamenco loudly, and Storage Net quietly. With the scale these businesses
bring, data onboarding fees for clients have decreased significantly to drive more engaged
use. We see this as an important competitive development, relative to cloud service
incumbents like AWS that offer extremely low barriers to initially store data (with significant
subsidies) but charge high fees on retrieval on the back end.

Trust is also important to address. One of Filecoin’s core innovations is verified storage:
All data stored on Filecoin’s network is verifiable by the client at any given time which
should make for a higher trust solution. However, blockchains and crypto are still broadly
misunderstood. Incidents like FTX’s bankruptcy have continued to contribute to a
perception around lack of safety with these platforms; many corporations have blanket
policies not to work with crypto-enabled platforms for fear of regulatory retribution.
Like the scaling of technological architecture mentioned above, the trust issue is also one
that will inevitably be solved by time. While politics can be uniquely tribal, technology
is not—over time the market will be able to differentiate between issues caused by human
action (FTX fraud, hacking accounts) and technology.

Utilization remains a first-principles driver of growth, but market growth is more

important. The real question is whether the Filecoin Virtual Machine, which brings
compute-and-training capabilities, will allow builders to access new primitives in the form
of compute-over-data (smart contracts on Filecoin). Conceptually, this is a return to the
capabilities of on-premises server storage. Compute-over-data allows for the processing
of computation to occur directly where the data is stored, vs. being transferred to a
central location for processing. This is an important limitation for projects using big
datasets that are compute power intensive, as latency creates massive cost and speed
impediments. As the field of AI scales, driven by underlying LLMs trained on data,
resource efficiency will become a bigger problem, and model training could become
prohibitively expensive for startups. Today LLM projects consume enormous amounts of
resources inefficiently training models with trillions of parameters over cloud servers,
not to mention undergo critical security events with each data transfer. With Filecoin’s
compute-over-data architecture, models can be trained and deployed locally, leveling the
playing field for smaller players trying to compete with behemoths. Beyond just
competitive cost, ecosystem participants expect that direct integrations with Filecoin’s
compute layers will drive the majority of building activity going forward, as proximity
to the data layer allows developers to create contextual products that are impossible
without composable software. Filecoin is facilitating smart contracts driven by private data.
At an extreme, and combined with encryption tech coming to market, Filecoin offers
the possibility that apps will be able to compute over people’s private data, without ever
revealing the actual contents of the data, creating brand new identity primitives and the
acceleration of permissionless commerce. Jonathan Victor, cofounder of Ansa Research,
had this to say about the transformation:

“Compute-over-data and the launch of FVM is Filecoin’s steam engine moment and will
allow developers to use stored data in dynamic ways to create brand new services and
primitives. FVM makes Filecoin the destination for any project experimenting with
AI-driven insights from large datasets. It’s not even going to be close on a cost basis versus
centralized providers of storage and compute.”

10 The next frontier in cloud computing: Decentralization as a catalyst for disruption

 Some of the early projects are building new capabilities for SPs in the Filecoin ecosystem.
Using Filecoin’s smart contract capabilities, a team has created a primitive for SPs to
borrow $FIL to fulfill their token leasing obligations. Others, like IoNet and Bacalhau are
creating GPU networks directly on Filecoin, leveraging the extensive resources the
network already has. Compute-over-data will expand the range of what the protocol can
be used for greatly, the result of which should be significantly more hosted and stored
data on the network. Exciting new use cases needs to be a priority, as the network will
need >10 times revenue to breakeven on an illustrative EBITDA basis.

As demand for cloud services surges, it is anticipated that scaling Filecoin’s compute
infrastructure will prompt a swift and competitive response from the industry. Unlike
traditional industries with simple corporate structures, differences between Filecoin’s
decentralized architecture and AWS/Azure’s centralized structure will take traditional
mergers off the table. This leaves two options more likely than any other:

1. Price competition: Given Filecoin’s cost structure is flexible and driven by non-cash
token emissions, price wars would just lead to significantly deteriorating margins for
incumbents making them unlikely to engage
2. Collaboration: More likely as incumbents can use the Filecoin network, and capture
pricing differences themselves

Collaboration is a more likely outcome than price competition: structural cost advantages
of Filecoin make it infeasible for big tech giants to compete at scale. We note how
the earliest disrupted incumbents in crypto have responded to the challenge to global
payments driven by stablecoins. Notably, incumbents such as PayPal13 understand the
value proposition as so differentiated as to risk regulatory interference in their core
business to participate. It makes far more sense to collaborate with crypto incumbents
(in this case Paxos) to drive better products for consumers.

A very similar situation to Filecoin, one can look at what’s happening in the telecom/
bandwidth space, where a similar idea, Nova Labs has already struck an agreement to use
Helium14 with a national incumbent (T-Mobile) to support the T-Mobile network, and
another with Telefonica for international roaming. Facing a very similar paradigm around
cost disadvantage, T-Mobile decided that leveraging Helium to offer collective phone
plans was the most economically efficient move. Helium Mobile has since launched
its service nationwide and is able to price their unlimited data phone plan at ~$20/month
(vs. competitors who charge more than 8 times that).15 With the Telefonica deal,
Helium Mobile will additionally be able to offer their low cost offering in Mexico, and allow
Telefonica clients to roam onto their US network.

The next few years will yield several partnerships, as incumbents realize the cost advan-
tages of these networks can be leveraged for their own use. Customers, at the end
of the day, are still king, and big tech has a better stranglehold on its customers than any
set of companies in history. The growth of these networks also gives some hope to other
big tech incumbents without cloud services engines; no doubt Apple, Meta, and others
will do everything they can to break the Amazon-Azure duopoly, especially as the power of
that infrastructure begins to affect their ability to compete. Collaboration will likely be
the long-term outcome as cloud-based SPs can use Filecoin as backend infrastructure
and capture the spread from providing storage directly to customers. Doing so can be

11 The next frontier in cloud computing: Decentralization as a catalyst for disruption

 extremely economically beneficial for them, as they turn their own fixed cost model into
a variable cost one. Filecoin’s biggest problems to date have been around scaling UX:
while storage is a commodity, the user experience is not. This is where big tech incum-
bents can significantly accelerate usage forward, and capture the entire price spread,
effectively capturing a >$20 spread on each terabyte of data (Exhibit 10).

Exhibit 10: Storage Cost Monthly Storage Costs Per Terabyte

25
Comparison $23.00

20

15

10

5 $4.00

$0.19 $0.47 $0.54

0
Filecoin Arweave Sia Storj Amazon S3

Source: Kassab, Sami; Bloomber, Seth; Grigore, Mihai. “The Essential Guide to Decentralized Storage Networks.” Messari. January 12, 2023.

DePIN truly has the ability to eat infrastructure the way software ate most of commerce.
Cloud services is a winner take all market, as the end resource is often commoditized and
therefore the markets depend on who can sustain the lowest price for the resource they
offer. This goes to production cost, and areas that DePIN wins are those where incentivized
communities can greatly lower production cost by contributing excess capital/labor like
spokes around a central hub of revenue-driving infrastructure. The fact that they are
winner-take-all markets further justifies the win-at-all-cost strategy that token incentiviza-
tion champions. That token incentivization can then drive a MOAT. The Lindy effect
holds as true in crypto as most internet-native markets. One of the lessons we’ve learned
the hard way with DePIN is that the first company to earn real revenue gets way ahead,
because that revenue is useful to sustain the network and back the token in the early
days by supporting SPs. At minimum scale, token value accrual is a big barrier to entry,
as once the flywheel is going, the token allows the network to borrow forward resources in
ways that allow faster growth through incentivization of the right partnerships and
suppliers in equity-like form. Getting this token economy going allows you to incentivize
at scale without a centralized step. As an example, Uber scaled significantly by raising
billions centrally and distributing cash as incentives to drivers. DePIN models would
allow direct incentivization of the end party which gives the network unbelievable access
to the end labor supply and allows companies to experiment with equity incentivization
in contribution for work, which we believe is a stickier and fairer prospect for storage
providers. Other upstarts continue to iterate on this model. Moving away from
Filecoin’s programmatic token incentive structure, Helium and Nova Labs have begun
experimenting with targeted incentivization, driving the ability to price discriminate:
varying incentives based on service quality. By specifically incentivizing the highest
value service, these dynamic models let networks better optimize exactly who they deliver
incentives to. We suspect over time Filecoin’s token structure may evolve to include
this powerful mechanism.

12 The next frontier in cloud computing: Decentralization as a catalyst for disruption

 Filecoin has yet to cross the chasm from a differentiated, loud contender to an all-encom-
passing protocol driving computation, but its strength in providing digital commodities
on premises is a significant driver of long-term potential. The network has sizable
embedded operating leverage, and scale in demand will reverberate meaningfully through
the system. As compute-over-data proliferates as a more powerful standard of data
management than core cloud, and as oligopolies move to clunky IAAS models, we expect
attention and revenue on Filecoin to accelerate meaningfully. Digital infrastructure,
despite being a public utility, is heavily concentrated in the hands of a small number of
companies because of massive scale advantages and regulatory capture. This has
welfare consequences to citizens, who have come to rely crucially on the internet in their
day-to-day lives and would suffer from oligopolistic king-makers driving the internet’s
future. Technology should be a rising tide for all boats by radically lowering delivery costs
and dramatically increasing inclusivity. While the first phase of the internet, structured
around open protocols, began to fulfill this promise, the second saw gatekeepers erect
high walls to charge ever higher tolls. DePIN is a radical way that the Internet can return
value to communities and users. By allowing people to own and improve the infrastructure
around them, DePIN spurs software-enabled collaboration at a scale that has only
previously been available to corporations and governments. DePIN could offer a unique
solution to escalating questions around resource productivity and welfare. Digital infra-
structure’s importance is only going to grow, and involving communities in their supply is a
powerful way to increase equity. With minimal disruption to their lives, people can create
tons of economic value. From education to remote work, the internet penetrates an
increasing proportion of how we live, learn, and grow. Should Filecoin’s model scale to
provide useful storage coverage, it may represent a capitalist example of how returning the
means of production to workers can create massive economic value.

Endnotes

1. Hamilton, B. (2023, November 15). The Clever Crafting of Microsoft’s Open AI Deal. Medium. Available at: https://medium.com/@brooks.hamilton/the-clever-crafting-of-microsofts-
open-ai-deal-4b54eaf0be86
2. ARK Invest. (2022, November 18). PaaS: The Goldilocks of IaaS and SaaS. ARK Invest. Available at: https://ark-invest.com/articles/analyst-research/paas-goldilocks-iaas-saas/
3. ThoughtWorks. (2021, November 21). Decentralized Security. ThoughtWorks. Available at: https://www.thoughtworks.com/en-us/insights/decoder/d/decentralized-security
4. W3Academy. (2024). Is Decentralized Cloud Storage the Next Big Web3 Industry? W3Academy. Retrieved from: https://w3academy.io/is-decentralized-cloud-storage-the-next-big-
web3-industry/
5. Fusion Worldwide. (2024). The Release of Filecoin Could Mean Product Shortages in the Semiconductor Industry. Fusion Worldwide. Retrieved from: https://info.fusionww.com/blog/
the-release-of-filecoin-could-mean-product-shortages-in-the-semiconductor-industry/
6. Destor. (2024). Filecoin Network Client Explorer. Destor. Retrieved from: https://destor.com/en-us/filecoin-network-client-explorer/
7. Filecoin’s protocol revenue represents the sum of: Base fees—Determined by blockspace congestion and required by any storage proof. Batch fees—Used for bundling storage proofs.
Overestimation fees – Required to optimize gas usage. Penalty fees—Collected for storage provider failures.”
8. Messari. (2023, January 15). State of Filecoin Q4 2023. Messari. Available at: https://messari.io/report/state-of-filecoin-q4-2023
9. Dixon, Chris. (2024). Read Write Own. Random House. p 131.
10. Messari. (2023, January 15). State of Filecoin Q4 2023. Messari. Available at: https://messari.io/report/state-of-filecoin-q4-2023
11. Amazon. (2024). Form 10-K. BAMSEC. Retrieved from: https://www.bamsec.com/filing/101872424000008?cik=1018724
12. https://kenstechtips.com/index.php/download-speeds-2g-3g-and-4g-actual-meaning
13. PayPal. (2023, November 15). PayPal Cryptocurrency—PYUSD. PayPal. Available at: https://www.paypal.com/us/digital-wallet/manage-money/crypto/pyusd
14. CoinDesk. (2022, November 15). Nova Labs says T-Mobile to cover 5G dead spots in Helium network. CoinDesk. Available at: https://www.coindesk.com/markets/2022/09/20/
nova-labs-says-t-mobile-to-cover-5g-dead-spots-in-helium-network/
15. HelloHelium. (Date, November 15). Hi5 Unlimited. HelloHelium Blog. Available at: https://blog.hellohelium.com/hi5unlimited/

13 The next frontier in cloud computing: Decentralization as a catalyst for disruption

WHAT ARE THE RISKS?
All investments involve risks, including possible loss of principal.
Blockchain and cryptocurrency investments are subject to various risks, including inability to develop digital asset applications or to capitalize
on those applications, theft, loss, or destruction of cryptographic keys, the possibility that digital asset technologies may never be fully
implemented, cybersecurity risk, conflicting intellectual property claims, and inconsistent and changing regulations. Speculative trading in
bitcoins and other forms of cryptocurrencies, many of which have exhibited extreme price volatility, carries significant risk; an investor can lose
the entire amount of their investment. Blockchain technology is a new and relatively untested technology and may never be implemented to a
scale that provides identifiable benefits. If a cryptocurrency is deemed a security, it may be deemed to violate federal securities laws. There may
be a limited or no secondary market for cryptocurrencies.
Digital assets are subject to risks relating to immature and rapidly developing technology, security vulnerabilities of this technology, (such as
theft, loss, or destruction of cryptographic keys), conflicting intellectual property claims, credit risk of digital asset exchanges, regulatory
uncertainty, high volatility in their value/price, unclear acceptance by users and global marketplaces, and manipulation or fraud. Portfolio
managers, service providers to the portfolios and other market participants increasingly depend on complex information technology and
communications systems to conduct business functions. These systems are subject to a number of different threats or risks that could adversely
affect the portfolio and their investors, despite the efforts of the portfolio managers and service providers to adopt technologies, processes and
practices intended to mitigate these risks and protect the security of their computer systems, software, networks and other technology assets, as
well as the confidentiality, integrity and availability of information belonging to the portfolios and their investors.
Any companies and/or case studies referenced herein are used solely for illustrative purposes; any investment may or may not be currently held
by any portfolio advised by Franklin Templeton. The information provided is not a recommendation or individual investment advice for any
particular security, strategy, or investment product and is not an indication of the trading intent of any Franklin Templeton managed portfolio.

IMPORTANT LEGAL INFORMATION

This material is intended to be of general interest only and should not be construed as individual investment advice or a recommendation or
solicitation to buy, sell or hold any security or to adopt any investment strategy. It does not constitute legal or tax advice. This material may not be
reproduced, distributed or published without prior written permission from Franklin Templeton.
The views expressed are those of the investment manager and the comments, opinions and analyses are rendered as of the publication date and
may change without notice. The underlying assumptions and these views are subject to change based on market and other conditions and
may differ from other portfolio managers or of the firm as a whole. The information provided in this material is not intended as a complete analysis
of every material fact regarding any country, region or market. There is no assurance that any prediction, projection or forecast on the economy,
stock market, bond market or the economic trends of the markets will be realized. The value of investments and the income from them can go
down as well as up and you may not get back the full amount that you invested. Past performance is not necessarily indicative nor a guarantee
of future performance. All investments involve risks, including possible loss of principal.
Any research and analysis contained in this material has been procured by Franklin Templeton for its own purposes and may be acted upon in
that connection and, as such, is provided to you incidentally. Data from third-party sources may have been used in the preparation of this material
and Franklin Templeton (“FT”) has not independently verified, validated or audited such data. Although information has been obtained from
sources that Franklin Templeton believes to be reliable, no guarantee can be given as to its accuracy and such information may be incomplete or
condensed and may be subject to change at any time without notice. The mention of any individual securities should neither constitute nor be
construed as a recommendation to purchase, hold or sell any securities, and the information provided regarding such individual securities
(if any) is not a sufficient basis upon which to make an investment decision. FT accepts no liability whatsoever for any loss arising from use of this
information and reliance upon the comments, opinions and analyses in the material is at the sole discretion of the user.
Products, services and information may not be available in all jurisdictions and are offered outside the U.S. by other FT affiliates and/or their
distributors as local laws and regulation permits. Please consult your own financial professional or Franklin Templeton institutional contact for
further information on availability of products and services in your jurisdiction.

Brazil: Issued by Franklin Templeton Investimentos (Brasil) Ltda., authorized to render investment management services by CVM per Declaratory Act n. 6.534, issued on October 1, 2001. Canada:
Issued by Franklin Templeton Investments Corp., 200 King Street West, Suite 1500 Toronto, ON, M5H3T4, Fax: (416) 364-1163, (800) 387-0830, www.franklintempleton.ca. Offshore Americas:
In the US, this publication is made available by Franklin Templeton, One Franklin Parkway, San Mateo, California 94403-1906. Tel: (800) 239-3894 (USA Toll-Free), (877) 389-0076 (Canada Toll-Free),
and Fax: (727) 299-8736. US: Franklin Templeton, One Franklin Parkway, San Mateo, California 94403-1906, (800) DIAL BEN/342-5236, franklintempleton.com. Investments are not FDIC insured;
may lose value; and are not bank guaranteed.
Issued in Europe by: Franklin Templeton International Services S.à r.l. – Supervised by the Commission de Surveillance du Secteur Financier – 8A, rue Albert Borschette, L-1246 Luxembourg –
Tel: +352-46 66 67-1, Fax: +352-46 66 76. Poland: Issued by Templeton Asset Management (Poland) TFI S.A.; Rondo ONZ 1; 00-124 Warsaw. South Africa: Issued by Franklin Templeton Investments SA
(PTY) Ltd, which is an authorized Financial Services Provider. Tel: +27 (21) 831 7400, Fax: +27 (21) 831 7422. Switzerland: Issued by Franklin Templeton Switzerland Ltd, Stockerstrasse 38, CH-8002
Zurich. United Arab Emirates: Issued by Franklin Templeton Investments (ME) Limited, authorized and regulated by the Dubai Financial Services Authority. Dubai office: Franklin Templeton,
The Gate, East Wing, Level 2, Dubai International Financial Centre, P.O. Box 506613, Dubai, U.A.E. Tel: +9714-4284100, Fax: +9714-4284140. UK: Issued by Franklin Templeton Investment Management
Limited (FTIML), registered office: Cannon Place, 78 Cannon Street, London EC4N 6HL. Tel: +44 (0)20 7073 8500. Authorized and regulated in the United Kingdom by the Financial Conduct Authority.
Australia: Issued by Franklin Templeton Australia Limited (ABN 76 004 835 849) (Australian Financial Services License Holder No. 240827), Level 47, 120 Collins Street, Melbourne, Victoria 3000.
Hong Kong: Issued by Franklin Templeton Investments (Asia) Limited, 62/F, Two IFC, 8 Finance Street, Central, Hong Kong. Japan: Issued by Franklin Templeton Investments Japan Limited.
Korea: Issued by Franklin Templeton Investment Trust Management Co., Ltd., 3rd fl., CCMM Building, 12 Youido-Dong, Youngdungpo-Gu, Seoul, Korea 150-968. Malaysia: Issued by Franklin
Templeton Asset Management (Malaysia) Sdn. Bhd. & Franklin Templeton GSC Asset Management Sdn. Bhd. This document has not been reviewed by Securities Commission Malaysia. Singapore:
Issued by Templeton Asset Management Ltd. Registration No. (UEN) 199205211E and Legg Mason Asset Management Singapore Pte. Limited, Registration Number (UEN) 200007942R.
Legg Mason Asset Management Singapore Pte. Limited is an indirect wholly owned subsidiary of Franklin Resources, Inc. 7 Temasek Boulevard, #38-03 Suntec Tower One, 038987, Singapore.
Please visit www.franklinresources.com to be directed to your local Franklin Templeton website.

US-GTM-3111356-3011356
© 2024 Franklin Templeton. All rights reserved. CLCLO-0724