Dataset · Q1 and Q2 2026 · vendor claim plus computed intervals

Backblaze model AFR with 95% intervals, Q1 and Q2 2026

Published Q1 and Q2 2026 Backblaze AFRs for 33 hard-drive models, each with an exact 95% Poisson interval computed from the published failures and drive days. Sample: Q1 2026 (January 1 to March 31) 30,203,180 drive days and 1,030 failures; Q2 2026 (April 1 to June 30) 31,553,350 drive days and 1,498 failures. Of 31 models in both quarters, the intervals do not overlap for 4 and overlap for 27. The Poisson assumption is a modelling choice made here, not something Backblaze publishes.

Download CSV33 rows · 14 columns · CSV

File /data/datasets/backblaze-2026-afr-confidence-intervals.csv · JSON metadata · Chart (SVG) · Model-level comparison · All datasets · AFR glossary entry

Population and method

Population: hard drives in Backblaze’s own data centers that met the report inclusion rule, in the Q1 2026 table (January 1 to March 31, 2026, 33 models) and the Q2 2026 table (April 1 to June 30, 2026, 31 models). Sample: Q1 has 30,203,180 drive days and 1,030 failures; Q2 has 31,553,350 drive days and 1,498 failures. Drive days, failures and AFR are copied from the printed tables.

The interval columns are computed here. For f failures the exact two-sided 95% Poisson interval (Garwood) is taken on the failure count and scaled by 365 × 100 ÷ drive days. This treats failures as a Poisson process over drive days with a constant rate per model and quarter. Backblaze does not publish intervals and does not state that assumption. The overlap column is “yes” when the Q1 and Q2 intervals share any value. Overlap is a cautious screen, not a hypothesis test.

Chart

Interval plot of Q2 2026 AFR for 31 drive models, each a dot at the published AFR with a bar for the exact 95% interval. Widest: ST16000NM000J, 0.06% to 14.30%. — Q2 2026 only. Every dot and bar end is a cell of the table below; the axis is capped at 20%.

Counted here

31 models are in both quarters. For 27 the 95% intervals overlap, for 4 they do not: HUH721212ALN604, ST8000NM0055, MG07ACA14TA, WUH721414ALE6L4. In all 4the Q2 rate is the higher one.
8 of 31 Q2 models had fewer than 10 failures. Their intervals are wide. The widest is ST16000NM000J at 0.06% to 14.30% around a published 2.57%.
A model with zero failures still has an upper bound above zero: 3.69 failures scaled to its drive days at 95%. A published AFR of 0 does not mean the true rate is zero.

Limits

One operator’s fleet, not an industry sample. The AFRs are vendor claims.
The Poisson model ignores drive age, batch effects, and failures that cluster in time, so real uncertainty is probably larger than shown. Association only.
31 overlap checks were made without a multiple-comparison correction. A non-overlap here is a flag to look, not a finding about cause.
Drive days enter as published. Models that grew or shrank during the quarter are weighted by their drive days, which is what the AFR already does.

Model table

Same 33 rows as the CSV. Empty cells mean the model is not in that quarter’s table. Intervals are 95%, in percent.
MFG	Model	TB	Q1 drive days	Q1 fail	Q1 AFR	Q1 low	Q1 high	Q2 drive days	Q2 fail	Q2 AFR	Q2 low	Q2 high	Overlap
HGST	HMS5C4040BLE640	4	15,858	0	0	0.00	8.49
HGST	HUH728080ALE600	8	86,921	0	0	0.00	1.55
HGST	HUH721212ALE600	12	234,350	8	1.25	0.54	2.46	236,107	7	1.08	0.44	2.23	yes
HGST	HUH721212ALE604	12	1,183,774	86	2.65	2.12	3.27	1,190,182	109	3.34	2.74	4.03	yes
HGST	HUH721212ALN604	12	870,252	95	3.98	3.22	4.87	861,297	180	7.63	6.55	8.83	no
Seagate	ST8000DM002	8	639,473	25	1.43	0.92	2.11	563,210	38	2.46	1.74	3.38	yes
Seagate	ST8000NM000A	8	21,552	1	1.69	0.04	9.44	21,499	0	0	0.00	6.26	yes
Seagate	ST8000NM0055	8	1,176,619	39	1.21	0.86	1.65	1,177,541	111	3.44	2.83	4.14	no
Seagate	ST10000NM0086	10	86,667	11	4.63	2.31	8.29	86,092	22	9.33	5.85	14.12	yes
Seagate	ST12000NM0007	12	87,389	11	4.59	2.29	8.22	87,496	10	4.17	2.00	7.67	yes
Seagate	ST12000NM0008	12	1,664,558	129	2.83	2.36	3.36	1,671,057	145	3.17	2.67	3.73	yes
Seagate	ST12000NM000J	12	96,887	1	0.38	0.01	2.10	97,746	0	0	0.00	1.38	yes
Seagate	ST12000NM001G	12	1,185,545	33	1.02	0.70	1.43	1,193,856	44	1.35	0.98	1.81	yes
Seagate	ST14000NM000J	14	37,258	1	0.98	0.02	5.46	43,349	0	0	0.00	3.11	yes
Seagate	ST14000NM001G	14	944,400	21	0.81	0.50	1.24	949,396	39	1.5	1.07	2.05	yes
Seagate	ST14000NM0138	14	111,709	15	4.9	2.74	8.08	110,410	25	8.26	5.35	12.20	yes
Seagate	ST16000NM000J	16	10,117	1	3.61	0.09	20.10	14,221	1	2.57	0.06	14.30	yes
Seagate	ST16000NM001G	16	3,098,105	44	0.52	0.38	0.70	3,162,117	56	0.65	0.49	0.84	yes
Seagate	ST16000NM002J	16	41,931	0	0	0.00	3.21	42,086	4	3.47	0.95	8.88	yes
Seagate	ST24000NM002H	24	820,090	74	3.29	2.59	4.13	861,587	78	3.3	2.61	4.12	yes
Toshiba	MG07ACA14TA	14	3,343,724	94	1.03	0.83	1.26	3,368,924	179	1.94	1.67	2.25	no
Toshiba	MG07ACA14TEY	14	88,602	5	2.06	0.67	4.81	89,164	6	2.46	0.90	5.35	yes
Toshiba	MG08ACA16TA	16	3,548,250	102	1.05	0.86	1.27	3,614,263	100	1.01	0.82	1.23	yes
Toshiba	MG08ACA16TE	16	553,878	21	1.38	0.86	2.12	568,379	16	1.03	0.59	1.67	yes
Toshiba	MG08ACA16TEY	16	435,063	26	2.18	1.42	3.20	439,243	44	3.66	2.66	4.91	yes
Toshiba	MG09ACA16TE	16	36,988	1	0.99	0.02	5.50	68,429	2	1.07	0.13	3.85	yes
Toshiba	MG10ACA20TE	20	1,665,263	42	0.92	0.66	1.24	2,117,524	64	1.1	0.85	1.41	yes
Toshiba	MG11ACA24TE	24	495,387	3	0.22	0.05	0.65	824,903	13	0.58	0.31	0.98	yes
WDC	WUH721414ALE6L4	14	776,557	7	0.33	0.13	0.68	783,318	23	1.07	0.68	1.61	no
WDC	WUH721816ALE6L0	16	267,868	24	3.27	2.10	4.87	268,875	30	4.07	2.75	5.81	yes
WDC	WUH721816ALE6L4	16	2,402,732	64	0.97	0.75	1.24	2,432,374	77	1.16	0.91	1.44	yes
WDC	WUH722222ALE6L4	22	3,992,942	42	0.38	0.28	0.52	4,146,339	63	0.55	0.43	0.71	yes
WDC	WUH722626ALE6L4	26	182,471	4	0.8	0.22	2.05	462,366	12	0.95	0.49	1.65	yes

Columns

mfg (string): Manufacturer label as published by Backblaze.
model (string): Drive model identifier as published.
size_tb (integer): Nominal capacity in terabytes as published.
q1_drive_days (integer): Drive days in Q1 2026 (January 1 to March 31), as published. Empty if the model is not in that table.
q1_failures (integer): Failures in Q1 2026 (January 1 to March 31), as published.
q1_afr_percent (number): AFR in percent for Q1 2026 (January 1 to March 31), copied as published.
q1_afr_low95 (number): Lower bound of the exact two-sided 95% Poisson (Garwood) interval for the AFR, in percent, computed here from failures and drive days, Q1 2026 (January 1 to March 31).
q1_afr_high95 (number): Upper bound of the same interval, in percent, Q1 2026 (January 1 to March 31).
q2_drive_days (integer): Drive days in Q2 2026 (April 1 to June 30), as published. Empty if the model is not in that table.
q2_failures (integer): Failures in Q2 2026 (April 1 to June 30), as published.
q2_afr_percent (number): AFR in percent for Q2 2026 (April 1 to June 30), copied as published.
q2_afr_low95 (number): Lower bound of the exact two-sided 95% Poisson (Garwood) interval for the AFR, in percent, computed here from failures and drive days, Q2 2026 (April 1 to June 30).
q2_afr_high95 (number): Upper bound of the same interval, in percent, Q2 2026 (April 1 to June 30).
intervals_overlap (string): yes if the Q1 and Q2 95% intervals share any value, no if they do not. Empty if the model is missing from either quarter. Computed here.

License

Derivative table transcribed from Backblaze Drive Stats reports for Q1 2026 and Q2 2026. Cite Backblaze if you use it. You are solely responsible for how you use it. Backblaze's dataset page allows sale of derivative works and forbids selling the Drive Stats data itself. This CSV is a small derived table with computed intervals, not Backblaze's daily SMART archive, and that archive is not for sale.

Backblaze Drive Stats use terms · Q1 2026 report · Q2 2026 report

Sources

Backblaze Drive Stats for Q1 2026 (accessed 2026-10-09)
Backblaze Drive Stats for Q2 2026 (accessed 2026-10-09)
Hard Drive Reliability & Test Data (accessed 2026-10-09)
Garwood, F. (1936). Fiducial limits for the Poisson distribution. Biometrika 28(3/4), 437-442 (accessed 2026-10-10)