https://eureka.patsnap.com/resources/blog/rd-blog/low-precision-training-formats-patent-landscape/ · Patsnap · data cut-off 2026-07-31 · downloaded from the live page
Patent Landscape · AI Training Infrastructure
Low precision training format patents: mapping block floating point, shared exponent and loss scaling claims
  • Concentrated at the top. The five leading assignees hold 41.8% of all 1,095 records in scope, and the top ten hold 59.9% — this is a field with a short list of entrenched filers, not a wide-open race.
  • Filing has cooled from its 2019 peak. 2019 was the high point at 115 filings; the complete-year trend shows 2021's 69 filings falling to 35 by 2024, a 49% drop over that span.
  • Compute and transmission overlap. G06F digital data processing covers 44.4% of records and G06N AI-model computing covers 19.2%, but H04B and H04L transmission classes each sit above 12%, showing format claims spilling into communications hardware.
Get a prior-art report on your approach
1,095
Published Records
42%
Top-5 Share of All Records
-49%
Filing Growth 2021→2024
US
Leading Jurisdiction

Filing growth compares 2021 (69 records) with 2024 (35) — a three-year span. 2024 is the most recent year we treat as complete: publication lags filing by roughly 18 months, so 2025 onwards are still filling in and any growth rate that ends there would understate the field. Top-5 share is the combined record count of the five largest assignees divided by all 1,095 records in scope (CR5), not by the ranked leaders only.

Published byPatsnap Research··8 min readSourced from Patsnap Eureka
Overview

What this landscape covers

Low precision training formats reduce the bit width used to represent weights, activations and gradients during neural network training — block floating point, shared exponent grouping, and the surrounding machinery of loss scaling, rounding mode selection and accumulation precision that keeps numerically unstable training runs converging. The dataset in scope spans 1,095 published records from 2015 through mid-2026, drawn from filings that combine format-level claims (block floating point, shared exponent) with the supporting numerical controls (dynamic range, loss scaling, accumulation precision, outlier channel handling, rounding mode, mantissa bit width).

Publication lags filing by roughly 18 months, so the most recent one to two years in any trend understate actual filing activity. Family counts, used throughout the ranking below, are the fairer unit for cross-jurisdiction comparison because they neutralise continuation filings and duplicate regional filings of the same invention.

Filing activity and technology composition, 2015–2026
  1. 1MICROSOFT TECHNOLOGY LICENSING LLC171
  2. 2APPLE INC89
  3. 3INTEL CORP82
  4. 4TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)69
  5. 5NVIDIA CORP47
  6. 6DOLBY LABORATORIES LICENSING CORP45
  7. 7ERICSSON INC44
  8. 8ALTERA CORP44
  9. 9WISTARIA TRADING INC38
  10. 10INTEGRATED DEVICE TECH INC27
Source: Patsnap Eureka. Assignee ranking and totals. Derived from a Patsnap search on Low Precision Training Formats covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
The data

Filing trend and technology composition

Two views of the same 1,095 records: how filing activity has moved year over year, and which IPC subclasses the claims actually sit in.

Filing trend, 2017–2026

Filings rose to a peak of 115 in 2019, then declined across complete-year data to 35 in 2024, a 49% drop from 2021's 69 filings. 2025 and 2026 figures are still filling in under the usual publication lag and should not be read as a continued decline.

Filing trend, 2017–202603060901203220172018115201920202021202220232024202542026Most recent year is partial — publication lag means later filings are not yet visible.

Technology composition by IPC subclass

G06F (electric digital data processing) and G06N (AI-model computing) anchor the field at 44.4% and 19.2% of the 1,095 records respectively. Four communications-related subclasses — H04B, H04L, H03M and H04N — each cover more than 9% of records, reflecting how format claims extend into transmission and coding hardware, not just training silicon.

Technology composition by IPC subclassG06F · Electric digital data processi…48644.4%G06N · Computing based on AI models21019.2%H04B · Transmission (general)13812.6%H04L · Digital information transmissi…13212.1%H03M · Coding & code conversion12511.4%G06T · Image data processing & genera…11810.8%H04N · Pictorial communication (video…1039.4%H04W · Wireless communication networks625.7%Other41938.3%

Shares are the percentage of the 1,095 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.

Source: Patsnap Eureka. Filing trend and technology composition. Derived from a Patsnap search on Low Precision Training Formats covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.

Go deeper on Low Precision Training Formats with Eureka

This page is one run against one query. Ask Eureka your own question about low precision training formats and every answer comes back with the patent numbers behind it.

Try Eureka
Key patents

Representative filing and most-cited prior art

Representative filing
US20190347072A12019-11-14

US20190347072A1 — Block floating point computations using shared exponents

MICROSOFT TECHNOLOGY LICENSING, LLC

A system for block floating point computation in a neural network receives a plurality of floating point numbers. An exponent value for an exponent portion of each floating point number is identified and mantissa portions of the floating point numbers are grouped. A shared exponent value of the grouped mantissa portions is selected according to the identified exponent values and then removed from the grouped mantissa portions to define multi-tiered shared exponent block floating point numbers. One or more dot product operations are performed on the grouped mantissa portions of the multi-tiered shared exponent block floating point numbers.Filed by Microsoft Technology Licensing, published 2019-11-14.

US20190347072A1 — patent drawing 1US20190347072A1 — patent drawing 2
View full filing
Most-cited records in this corpus
#Publication no.Patent titleCitations
1US5889868AOptimization methods for the insertion, protection, and detection of digital watermarks in digitized data846
2US6522767B1Optimization methods for the insertion, protection, and detection of digital watermarks in digitized data347
3US20120250740A1OFDM signal processing in a base transceiver system288
4US5537435ATransceiver apparatus employing wideband FFT channelizer with output sample timing adjustment and inverse FFT…233
5US20130322753A1Systems and methods for local tone mapping226
6US5839100ALossless and loss-limited compression of sampled data signals216
7US4760525AComplex arithmetic vector processor for performing control function, scalar operation, and set-up of vector s…215
8US20190340492A1Design flow for quantized neural networks210
9US5410750AInterference suppressor for a radio receiver210
10US20040012600A1Scalable high performance 3d graphics197

Citation counts favour older filings that have had more time to accumulate citations within the searched corpus — read them as a signal of influence on later filings, not as a ranking of current technical importance.

Each row carries its publication number; clicking a row searches Eureka by that number.

Source: Patsnap Eureka. Citation counts and representative records. Derived from a Patsnap search on Low Precision Training Formats covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
Run it yourself

Put your own technology through the same analysis


  
Where to run it
Fastest

Eureka on the web

When you want the answer in the next five minutes.

The agent works the prompt against patents and technical literature, citing every source.

Run your analysis now →
For builders

MCP server & REST API

When it has to run inside your own pipeline.

Patent search, landscape analysis and assignee resolution as MCP tools. Drop them into any agent framework, or call REST directly.

Browse MCP servers →
Insights

What the numbers mean for filing strategy

Three patterns worth reading before deciding where to file or challenge.

Concentration
41.8% / 59.9%
share of 1,095 records held by top 5 / top 10 assignees

A short list of entrenched filers

The five leading assignees combined hold 41.8% of all 1,095 records in scope, rising to 59.9% for the ten leading assignees. That leaves a long tail of single- or low-count filers competing for the remaining claim space — a structure typical of a field where the core numerical techniques were staked out early by a handful of large compute and communications incumbents.

Based on the 100-company assignee ranking returned by the dataset.
Momentum
-49%
2021 (69 filings) to 2024 (35 filings), complete-year data

Filing has pulled back from its 2019 peak

Filing peaked at 115 in 2019 and has trended down since, with the last fully-published years showing a 49% drop from 2021 to 2024. Because publication lags filing by around 18 months, 2025-2026 figures are still incomplete and should not be read as confirming a continued slide — but the multi-year direction through 2024 is real.

2025 and 2026 filing counts will continue to rise as records publish.
Technology spread
44.4% / 19.2%
share of records in G06F / G06N

Format claims cross into transmission hardware

Electric digital data processing (G06F) and AI-model computing (G06N) dominate, but transmission subclasses H04B and H04L each exceed 12% of the 1,095 records. This overlap suggests block floating point and shared exponent techniques are being claimed for signal and video pipelines as well as neural network training, widening the prior art a new filer must clear.

Class shares sum above 100% because records can carry multiple IPC classes.
Eureka AI Agent
Looking for what nobody has claimed yet?

Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to low precision training formats, with the prior art for and against each one.

Find the white space →
Source: Patsnap Eureka. Co-assignee relationships and derived observations. Derived from a Patsnap search on Low Precision Training Formats covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
Players

Who holds the claim space

The leading assignee sits well above the rest of the ranked field, and recent-year filing activity has thinned out even among the incumbents.

Leader
171 records
leading assignee's family count

One filer well ahead of the field

The leading assignee holds 171 records, more than triple the fifth-place count of 47 and well above tenth place at 27. That gap at the top is the clearest structural feature of this dataset — it is not an evenly distributed field.

Counts are patent families, not raw publications.
Recent momentum
-80% to -100% YoY
latest-year change among several leading assignees

Even incumbents have slowed recently

Several of the most active historical filers show sharp year-over-year declines in the latest tracked year, with some down to zero and one down 80% year over year. Given the publication lag, this looks more like incomplete recent data than an actual retreat from the technology.

Momentum figures reflect the most recent tracked year, which is still filling in.
Collaboration
10 pairs
co-assignee pairs identified

Co-filing is limited and concentrated

Only ten co-assignee pairs appear in the dataset, and the strongest pairing appears 19 times — modest compared with the leading solo filer's 171 records. Most activity in this field is filed by a single assignee rather than through joint ventures or cross-licensing structures.

Pair counts reflect shared filings across the corpus.
🔍
Under-claimed branches worth checking before you file
These sub-areas sit inside the broader IPC composition but show thinner concentration at the top, based on the technology and class spread above.
Outlier channel isolation for activation quantizationRounding-mode selection for gradient accumulationMixed mantissa bit-width scheduling across layersDynamic loss-scaling control loopsShared-exponent grouping for video/signal pipelines
Rank all filers by momentum →
Recent-year filing momentum by assignee
AssigneeRecent yearYoY
Intel Corp1-80%
NVIDIA Corp1—
Microsoft Technology Licensing, LLC0-100%
Telefonaktiebolaget LM Ericsson (publ)0-100%
Wistaria Trading Inc0—
Apple Inc0—
Ericsson Inc0—
Dolby Laboratories Licensing Corp0—
Source: Patsnap Eureka. Assignee-level momentum. Derived from a Patsnap search on Low Precision Training Formats covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
What's next

Where to take this

Two directions depend on whether you are scoping freedom-to-operate or looking for filing opportunity.

Check claim scope before you file

The concentration at the top of this field means a new filing in block floating point or shared exponent grouping is likely to sit close to existing claims from the leading assignees. Reading the independent claims of the leading filer's family is a faster first step than a full landscape search.

Explore claims in Eureka

Track the under-claimed branches

Rounding-mode selection, mixed mantissa scheduling and outlier channel isolation show thinner concentration than the core block floating point claims. These are worth monitoring for new entrants before they consolidate the way the core format claims have.

Set up monitoring in Eureka
Source: Patsnap Eureka. Forward-looking reading of the same dataset. Derived from a Patsnap search on Low Precision Training Formats covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
FAQ

Common questions on low precision training patents

Answers are grounded in the same dataset. Derived from a Patsnap search on Low Precision Training Formats covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP

Research Low Precision Training Formats in depth with Eureka

Go past this page: query the whole low precision training formats corpus yourself, in your own scope.
Every answer comes back with patent numbers you can open.

Try Eureka

Disclaimer. This page is generated from Patsnap Eureka data drawn from a limited snapshot of global patent and scientific-literature records, and is provided for general information and reference only.

Patent data carries inherent limitations: recent filings (typically the most recent 18–24 months) are under-counted due to standard publication lag; counts may be reported at either a patent-family or a patent-record basis and are not always directly comparable; classification, applicant-name, and citation data may contain errors, duplicates, or omissions; and the underlying search query defines and constrains the scope shown. As a result, the analysis may be incomplete or inaccurate and may not reflect the full technology landscape.

Nothing on this page constitutes an exhaustive prior-art, novelty, freedom-to-operate, or validity search, nor does it constitute legal, financial, investment, or professional advice, and it should not be relied upon as such. Any patent, commercial, or strategic decision should be verified independently and reviewed with qualified patent, legal, and domain professionals. Patsnap makes no warranties, express or implied, as to the accuracy, completeness, or fitness for any particular purpose of the information presented.

Machine translation. Assignee and organisation names originally recorded in Chinese, Japanese or Korean have been rendered into English by an AI translation step so that the tables stay readable. These renderings are best-effort and may not match a company's registered English name; the original name is what the underlying patent record carries, and it is what any Eureka query launched from this page uses.

Help us improve

Something wrong, out of date, or missing on this page? Tell us and it goes straight to the team that maintains it.