How Much Data Is Created Around the World

By Muninder Adavelli. Scraping, SEO, and proxies for over a decade.

1. The top-line: 402 million terabytes a day, 181 zettabytes a year

1.1 The numbers that anchor every other stat

Statista's headline for 2024 is 402.74 million terabytes of data generated per day: the equivalent of 4.0274 × 10²⁰ bytes, encompassing new data plus captured, copied, and consumed data. That maps to 149 zettabytes for the full year. By 2025 the annual figure grows to 181 zettabytes (Rivery and ExplodingTopics both carry this number, tracing back to Statista). ExplodingTopics projects 221 zettabytes for 2026.

These are not competing forecasts. They are consecutive years on the same growth curve: roughly 22% year-over-year. IDC has separately reported that globally stored data doubles about every four years, and Rivery notes that approximately 90% of the world's data has been generated in just the past two years.

The disagreement in the field is not about the arithmetic. It is about what "created" means, which is the subject of the next few paragraphs.

1.2 The 20-year climb from less than 1 ZB to 181

The trajectory looks obvious in retrospect. Global storage capacity was 2.6 exabytes in 1986, 15.8 EB in 1993, 54.5 EB in 2000, 295 EB in 2007, 5,000 EB in 2014, and 6,800 EB by 2020. The world was generating less than one zettabyte per year roughly 20 years ago, and only reached 2 ZB per year in 2010. Fifteen years later, we are at 181 ZB annually: a roughly 90-fold jump.

The curve is not linear. The 2020-to-2025 leap is steeper than the whole preceding decade, and IoT, cloud services, and AI infrastructure are the reason.

1.3 What "data created" actually counts (and what it doesn't)

DesignRush, in its 2024 review, flags the definitional problem directly: there is no single measurement for daily data creation, and estimates may refer to new data generated, data copied and consumed, or data transmitted across networks. IDC's former senior vice president David Reinsel is the source of the most-quoted caveat: only about 10% of the data flowing through the datasphere is unique, newly created or captured information. The other 90% is copies, retransmissions, and reprocessing, with what Reinsel calls the "recursion rate" growing exponentially as the same data gets reused across more services.

That single fact means every "X zettabytes created" number in this article is inflated. Whether you count a CDN cache hit as "creation" is a choice. Statista and IDC choose to include it; the physical volume of unique new data in 2024 is closer to 15 zettabytes than 149.

2. What happens in a single day, in numbers

The abstract totals become tangible when you break them into activity counts. All of the following are drawn from OpenText's 2024 aggregation of secondary sources.

2.1 Communications: 361 billion emails, 27 billion texts, 8.5 billion Google searches

Text-based communications dominate the per-day activity totals. OpenText cites Statista for the 361.6 billion global email figure, EarthWeb for the 27 billion daily text messages, and DemandSage for the 8.5 billion Google searches per day. This is one of the parts of the "data created" story where numbers actually agree across sources: the aggregator pattern (one blog citing another citing the primary research firm) means the same underlying numbers propagate widely.

The volumes are per-day, not per-user. Global emails per person per day works out to about 45 emails; per internet user, closer to 68.

2.2 Media: 23 million TikTok videos, 522 million Netflix hours

Video is where per-user bytes actually pile up. TikTok uploads 23 million videos per day; Netflix streams 522 million hours. The 522M-hour figure is roughly 60,000 continuous years of viewing every 24 hours.

AppLogic Networks (the company formerly known as Sandvine) puts video at ~76% of mobile traffic in its 2025 Global Internet Phenomena Report. Their earlier Sandvine breakdowns give the fuller mix: social media at ~18% of downstream traffic (~2.6 GB per subscriber per day), gaming at ~3% (~398 MB per day), messaging at ~0.6% (~81 MB per subscriber per day). Cloud gaming, when it appears, hits ~22 GB per hour: an outlier that skews per-user averages upward wherever it takes hold.

2.3 AI: 2.5 billion ChatGPT prompts, 12.3 million Gemini visits, 34 million generated images

AI is the newest and fastest-growing category in the daily-activity mix. Axios, reported via TechCrunch in July 2025, put ChatGPT prompts at 2.5 billion per day: more than double a year earlier. Google Gemini receives more than 12.3 million visits per day. AI tools generate more than 34 million images daily, and organizations collectively spend nearly $1.8 billion per day on AI technology and services.

None of these numbers show up in Statista's or IDC's traditional data-creation totals; AI-generated content lands in the unstructured-data bucket rather than being isolated as its own line item.

3. The per-user picture: 21 GB a month per smartphone

3.1 Ericsson's Mobility Report: 21 GB per month per smartphone

Ericsson's 2025 Mobility Report is the most-defensible per-device number in circulation. Average smartphone traffic is 21 GB per month, ranging from the low-20s to the mid-30s of GB per month depending on region. This is a device-level measurement of everything a smartphone pushes and pulls: background syncs, app updates, streaming, ads, silent telemetry.

It is not a per-person creation figure. A single user with a phone, a laptop, and a smart TV crosses many of these device-level counters simultaneously. Ericsson's number tells you how much any given smartphone touches; it does not tell you how much data any given human causes to exist.

3.2 Video dominates the growth in per-user bytes

The reason per-device numbers keep climbing is video. AppLogic Networks' 2025 report shows video at ~76% of mobile traffic. A separate figure cited in DesignRush puts videos at ~82% of overall internet data traffic when fixed-line consumption is included. The composition matters because video is the highest per-byte category: a streamed hour of high-definition video costs orders of magnitude more bytes than an hour of messaging or a day of search queries.

The same content also gets counted multiple times as it moves through CDN caches, edge servers, and end devices, reinforcing the recursion problem.

3.3 Only about 10% is actually new

Reinsel's estimate is worth stating twice. When you see "402M terabytes created daily," you are seeing the cumulative footprint of every cache hit, every CDN fetch, every AI inference that regenerates the same data. Ericsson's per-device 21 GB/month is not creation either: it is consumption plus the reprocessing that happens on the way to consumption. The rate at which the same data is reprocessed is growing exponentially, and every copy inflates the total.

4. AI as a new data-creation engine

4.1 34 million images and $1.8 billion in daily spend

The 34-million-images-per-day and $1.8-billion-per-day AI-spending numbers both come from OpenText's 2024 aggregation of daily internet activity. These are current, measurable volumes, not projections. Every image is a new visual file, generated on demand by tools like Midjourney, DALL·E, and Stable Diffusion.

Neither number appears in Statista's global data-creation forecast or in IDC's zettabyte projections. That is a real blind spot in the top-line numbers: an entirely new category of measurable, recent, high-volume data creation is not being tracked in the metrics people cite when they answer "how much data is created?"

4.2 2.5 billion ChatGPT prompts a day, doubled year over year

Axios, reported via TechCrunch in July 2025, put ChatGPT prompts at 2.5 billion per day: more than double the prior year. Every prompt is a small data event: a user input, a model response, and, in most deployments, log entries for compliance, safety, and quality-of-service monitoring. The bytes per prompt are small; the aggregate cadence (29,000 prompts per second, on average) is the story.

4.3 AI output lands in the "unstructured data" bucket

DBMR's data-wrangling market report notes that the unstructured data segment is expected to witness the fastest CAGR from 2025 to 2032. AI-generated content is folded into this segment rather than counted separately. The consequence is straightforward: when a market analyst tells you unstructured data is growing fastest, they are partly describing AI output, but the connection is rarely made explicit, and no source in this pool breaks out the AI-specific share.

5. IoT and sensor data: 21 billion devices, generating continuously

5.1 21.1 billion IoT devices, growing 14% year over year

IoT Analytics counts 21.1 billion active IoT devices by the end of 2025, up 14% from 2024, with the total trending toward 39 billion by 2030. Each device is a small, continuous data source: temperature, motion, pressure, location, status. The device count is not the data volume, but it is the leading indicator; every added device increases the baseline production of always-on machine data.

5.2 Sensor data analytics: $18.5B today, $65.85B by 2033

DBMR's sensor data analytics market report puts the current value at $18.50 billion and projects $65.85 billion by 2033, implying a 17.20% CAGR from 2026 to 2033. North America dominated the market in 2025 with a 34.5% revenue share; Asia-Pacific is expected to be the fastest-growing region at 8.1% CAGR. The manufacturing segment held 26.5% of the market, and the software segment held 42.8%.

Market dollars and data volume are different metrics: a $65 billion analytics-tools market does not directly tell you sensor-data volume. But the direction is unambiguous: capital is flowing into sensor-data infrastructure faster than into most other data-adjacent categories in the DBMR pool.

5.3 Industrial data management: 11.2% CAGR, driven by Industry 4.0

The global industrial data management market was valued at $103 billion in 2024 and is projected to grow at an 11.20% CAGR through 2032, per DBMR. India is expected to see the highest CAGR in the segment, driven by government digitalization initiatives and manufacturing expansion. The category covers real-time monitoring, predictive analytics, and closed-loop automation across factories, energy grids, and supply chains: machine data generated by production processes, not by users.

6. Log data is the hidden explosion

6.1 Log management market: 13.1% CAGR

DBMR values the global log management market at $2.83 billion in 2024 and projects a 13.1% CAGR from 2025 to 2032. China is expected to see the highest regional CAGR. The market's growth is not driven by new user-facing content; it is driven by the requirement to record what already exists. Every API call, login, error, and system event generates a log line; scaling infrastructure means scaling the log volume that comes with it.

6.2 Cyber risk and compliance are the drivers

DBMR flags the growth drivers explicitly: increased cyber risk exposure, rising digital services, and stricter compliance requirements. Regulatory frameworks such as GDPR, HIPAA, and CCPA require audit trails, and cybersecurity investigations depend on log data. This is data that must be generated and retained whether or not it is ever read.

AWS reported a 31% year-over-year increase in log management and monitoring service usage as of March 2024, driven by enterprise migration to multi-cloud architectures: a supporting data point that log volume is not a projection but an active surge.

6.3 The SME cost and skill gap

DBMR flags two persistent barriers for small and medium enterprises: high implementation costs and integration complexity, plus a shortage of skilled cybersecurity professionals needed to actually use the collected logs. TechRepublic reported in January 2024 that over 47% of SMEs surveyed cited cost and integration difficulties as key barriers to adopting security and log management platforms.

The upshot is that log data gets generated at scale by large enterprises with the budgets to store and analyze it, while SMEs generate the same data but leave it unused. The volume grows either way.

7. Where all this data physically lives

7.1 $134 billion for next-generation storage

DBMR values the global next-generation data storage market at $68.98 billion in 2021 and $96.30 billion in 2025, projected to reach $134.45 billion by 2029 at an 8.70% CAGR. The category covers solid-state architectures, cloud-native data fabrics, and AI-optimized storage: the physical layer that absorbs the daily and annual totals cited earlier in this piece.

7.2 Customer data platforms: $85 billion by 2032, at 33.70% CAGR

The customer data platform (CDP) market (the tooling that stitches first-party customer data into a unified profile) was valued at $8.34 billion in 2024, is projected to reach $26.65 billion in 2028, and $85.18 billion by 2032. The 33.70% CAGR (2025 to 2032) makes it the fastest-growing data-tooling category in the DBMR pool covered here: outpacing sensor analytics (17.20%), log management (13.1%), and next-gen storage (8.70%).

North America dominates the CDP market, and Asia-Pacific is the fastest-growing region.

7.3 Data wrangling: 10.7% CAGR as data outstrips capacity to shape it

DBMR's data wrangling market was $3.32 billion in 2025 and is projected to reach $7.48 billion by 2033, growing at 10.70% CAGR. North America held the largest revenue share at 38.75% in 2025; Asia-Pacific is expected to grow fastest through 2033. Data wrangling is a useful proxy for a different problem: not creation, but readiness. The category measures how much money is being spent to make raw data usable, and its steady growth signals that data quality is not keeping pace with data volume.

8. Real-world data creation, from Walmart to offshore rigs

8.1 Retail and manufacturing: Walmart at 60 PB/day, Seagate at 80 TB/day per factory

DesignRush cites Walmart at roughly 2.5 petabytes per hour, or ~60 PB per day, spanning inventory tracking, RFID scans, supply chain telemetry, and video surveillance across thousands of stores. Seagate's factories, meanwhile, generate 50.5 petabytes total from July 2023 to March 2025 (averaging about 80 TB per day per facility) and inspect up to 3 million images per day for defect detection. Industrial operations broadly average around 80 TB per day per site.

These are per-operation numbers, not global aggregates. They matter because they illustrate scale that never appears in consumer-facing "data created" reports.

8.2 Energy and healthcare: 1-2 TB/day per offshore platform, 500 MB to 1 GB per MRI

Offshore oil platforms generate 1 to 2 TB of data per day each, from seismic sensors, drilling telemetry, and real-time pressure monitoring. A single MRI scan produces 500 MB to 1 GB of raw image data. Neither of these operations produces user-visible content, but both are continuously generating data that adds to the global total, and both are invisible in the DOMO-style "Data Never Sleeps" infographics that anchor most published treatments of this topic.

9. Automatic identification: barcodes, RFID, biometrics

DBMR's automatic identification and data capture market was valued at $84.85 billion in 2025, growing at a 21.55% CAGR from 2025 to 2033. Barcodes dominated with a 38.6% share; Asia-Pacific held 41.82% of the regional market. The RFID products segment is projected to grow fastest through 2033, driven by contactless tracking, smart inventory, warehouse automation, and real-time asset monitoring. North America is projected at 8.13% CAGR through 2033, driven by warehouse automation, healthcare digitization, and smart retail.

None of this data flows into "how much data is created" listicles either, but it represents billions of daily data-creation events: a barcode scan, an RFID pass, a biometric read.

10. What the numbers don't capture

10.1 Only 10% is new: everything else is a copy

Reinsel's estimate deserves a third mention because it is load-bearing. When you see "402M TB created daily" or "181 ZB annually," you are looking at the aggregate footprint of every copy, cache, replica, and re-encoding, not a count of unique creation events. The recursion rate is climbing, which means the gap between "data created" and "unique data created" widens each year.

If you strip the copies out, the 181 ZB figure for 2025 becomes something closer to 18 ZB. That is still an enormous number, but it is an order of magnitude smaller than the headline.

10.2 What is not in any of these numbers

The sources gathered here do not measure several categories the question "how much data is created" implicitly asks about. Deep-web and dark-web activity generates data that is not publicly indexed and not counted in any of the aggregators referenced above; no reliable estimate exists in the pool. Per-platform breakdowns for Facebook, Instagram, YouTube, and Discord are not present in the current source set: many secondary sources cite recycled 2019 to 2020 upload counts, but those numbers do not appear in the primary pool. There is no measurement of water use per gigabyte for creation (only for transmission and inference in unrelated reports). Per-country data-creation splits are absent; the DBMR reports give regional market shares for the tooling markets, but not for underlying data volumes.

Individual AI-provider volumes are largely missing too. Only ChatGPT (2.5B prompts/day) and Google Gemini (12.3M visits/day) surface in the pool as specific numbers. There are no comparable figures for Meta AI, Microsoft Copilot, Anthropic's Claude, or AWS Bedrock, despite each being a plausible participant in the daily data-generation totals.

The most defensible summary is the one nobody wants to publish: we can measure with reasonable confidence how much data flows through the visible parts of the datasphere, we can name the categories that grow fastest, and we can put honest numbers on physical operations like factories, oil platforms, and hospitals. What we cannot do is give one clean answer to the question the headline implies. The 402M-terabytes-a-day number is real. It is also inflated, incomplete, and mostly composed of copies of things that already existed.